Glean Release Notes

Release version: 559

This release makes Agent Identity generally available, brings the Fathom, Crayon, and Veeva Vault connectors and the Platform Skills API to general availability, and turns on the updated left navigation for everyone. It also adds six new Model Context Protocol (MCP) server templates, new admin controls for Salesforce API usage and voice transcription, and a set of reliability fixes.

Major launches

Agent Identity is now generally available

Give each Glean agent its own identity — its own governed account in every system it connects to — so agents act and appear as themselves instead of running under a person's credentials. Builders can attach admin-governed service credentials to their agents and configure dedicated Slack and Teams apps, while admins manage service credentials and Knowledge Profiles using existing access controls. The result is clearer attribution for agent activity and tighter governance over what each agent can reach.

How to access: Available by default. Admins manage service credentials and Knowledge Profiles in the Admin console; builders attach credentials when configuring their agents.

Learn more
Agents [ROAD-1542]

Fathom connector now generally available

The Fathom connector is now generally available. Connect Fathom to bring meeting recordings, transcripts, and summaries into Glean, so your team can search meeting content and ask Assistant about it alongside the rest of your company knowledge — with each person's Fathom permissions respected.

How to access: Admins set up Fathom from the Admin console under Data sources.

Learn more
Connectors [ROAD-1664]

Crayon connector now generally available

The Crayon connector is now generally available. Connect Crayon to index battlecards and AI-generated Sparks, making competitive intelligence insights searchable in Glean so sales and product teams can find the latest competitive context where they already work.

How to access: Admins add Crayon from the Admin console under Data sources, provide a Crayon API key, and start a crawl.

Learn more
Connectors [ROAD-1259]

Veeva Vault connector now generally available

The native Veeva Vault connector is now available to set up from the Add data source page. It indexes quality documents and related metadata while respecting each user's authorized access in Veeva Vault, so regulated content stays permissions-aware in Glean search and Assistant answers.

How to access: Admins configure Veeva Vault from the Add data source page using their Vault URL and integration credentials.

Learn more
Connectors [ROAD-1042]

Platform Skills API now generally available

The Skills API — 14 operations under /api/skills/ — is now generally available, so you can discover, validate, import, sync, version, and manage skills programmatically. Integrations can call these endpoints without sending the X-Glean-Include-Experimental header.

How to access: Available through the Glean Platform API with an appropriately scoped API token.

Learn more
Developer Platform [ROAD-1473]

Updated left navigation now on for everyone

The reorganized left navigation is now generally available. Navigation items are arranged into tiers so your most-used destinations stay visible while the rest move into an overflow menu. You can customize which items appear in your own sidebar, and admins can set the default sidebar layout for the organization.

How to access: On by default. Customize your own sidebar directly in the left navigation; admins set workspace defaults under Admin console → Appearance.

Learn more
Search experience [ROAD-1603]

Six new MCP server templates

The Model Context Protocol (MCP) server library keeps growing. This release adds six new remote server templates: GIS Cloud for maps, layers, and spatial data; Lightrun for inspecting live runtime values and metrics; MYOB for read-only financial queries; PaidSync for analyzing and managing advertising campaigns; Avoma for meeting tools; and Prisma Postgres for database management. Most templates authenticate with per-user OAuth.

How to access: Admins add and enable the servers they want from the MCP server directory in the Admin console.

Learn more
Actions [ROAD-1441]

Critical updates

SharePoint group sites discovered with split app permissions

Fixed SharePoint group site discovery for deployments where the Microsoft 365 identity app lacks site enumeration access but the SharePoint app has it. Group sites are now discovered using the app that holds the required permission, so site coverage stays complete.

How to access: No action needed. The fix applies automatically.

Connectors

Document scans no longer skip documents

Fixed a pagination issue that could omit documents during scans of sharded document storage, so scans now reliably cover every document.

How to access: No action needed. The fix applies automatically.

Platform

Enhancements

Control Glean's daily Salesforce API usage

Admins can configure Glean's daily Salesforce API-call budget directly in the Salesforce connector's Advanced settings, making it easier to manage API consumption alongside other integrations. The minimum setting is 16,500 calls per day.

How to access: Admin console → Salesforce connector → Advanced settings.

Learn more
Connectors [ROAD-1742]

Published SharePoint documents stay searchable during drafts

Glean now keeps the latest published version of a SharePoint document searchable while a newer draft version exists, so published content no longer disappears from search while someone works on an update.

How to access: No action needed. The behavior applies automatically to SharePoint content.

Learn more
Connectors [ROAD-1516]

User data masked in agent traces (beta)

Agent traces available to agent builders now mask user data, protecting personal information while builders review and debug agent runs. Masking applies to traces going forward.

How to access: Available in beta. Agent builders see masked traces automatically when viewing agent runs.

Agents [ROAD-1645]

Launched features grouped by date in Feature rollouts

The Launched tab on the admin console Feature rollouts page now groups features that became generally available on the same date into a single row and lists dates with the most recent first. Select the GA on column header to reverse the order.

How to access: Admin console → Feature rollouts → Launched tab.

Learn more
Admin Capabilities [ROAD-732]

Product type breakdown in billing CSV exports

Billing CSV exports now honor the Product type breakdown setting for department views and Glean usage group views, adding product-specific usage columns when the setting is enabled.

How to access: Turn on Product type breakdown in the billing dashboard, then export the CSV.

Admin Capabilities

Longer agent descriptions

Agent descriptions can now contain up to 2,000 characters, giving you more room to explain an agent's purpose and behavior so teammates know exactly what an agent does before they run it.

How to access: Edit the description when creating or updating an agent.

Agents

Find past agent runs with the Agents filter

Chat search now has an Agents filter, so you can find past agent runs without scrolling through all of your chats.

How to access: Open search from your chat history and select the Agents filter.

Learn more
Assistant [ROAD-1684]

Protection against overwriting a newer agent version

Glean now detects when an agent's published version changes while you are editing it. If you try to publish from an older version, you can save your work as a version or load the latest version before publishing again, helping prevent accidental overwrites.

How to access: No action needed. The check runs automatically when you publish an agent.

Agents

Google Drive file descriptions and custom properties indexed

Google Drive crawls now include file descriptions and custom properties in metadata, so this information can be used when indexing documents and improving search relevance.

How to access: No action needed. The metadata is included automatically in Google Drive crawls.

Connectors

Dedicated calendar skills for Google Calendar and Outlook

Calendar scheduling, rescheduling, and agenda creation now use dedicated Google Calendar and Microsoft Outlook skills, selected automatically according to your connected calendar tools, for more reliable meeting workflows.

How to access: No action needed. The right calendar skill is selected automatically.

Assistant

Set up Box MCP tools and indexing together

Admins can now set up Box MCP tools and Box indexing together in one guided connector setup flow, instead of configuring them separately.

How to access: Admin console → Connectors → Box.

Connectors [ROAD-941]

Consistent Slack access enforcement

The Slack unified connector now applies admin user-access selections consistently to both real-time and synced Slack content, so the people you select see the same Slack results everywhere.

How to access: No action needed. Existing user-access selections are enforced automatically.

Connectors

GPT Transcribe for realtime voice transcription

GPT Transcribe is now available as a selectable realtime voice transcription model for deployments that use their own OpenAI or Azure OpenAI key, on version 559.0 and later.

How to access: Admins using their own OpenAI or Azure OpenAI key can select GPT Transcribe as the realtime voice transcription model in the Admin console.

Admin Capabilities [ROAD-1743]

New usage scopes for API tokens and OAuth clients

API tokens and OAuth clients can now use two usage scopes. USAGE reads and evaluates usage for the signed-in user and agents they can edit. ADMIN_USAGE manages workspace usage limits and settings, and still requires billing permissions.

How to access: Select the scopes when creating or updating API tokens or OAuth clients.

Developer Platform

Clearer model picker icons

Model picker icons are updated for readability, including an effort indicator in the compact composer, so it's easier to compare models at a glance.

How to access: No action needed. The updated icons appear automatically in the model picker.

Assistant

Fixes

More reliable code execution in chat

Improved recovery from stalled code execution: command timeouts now return an error to the assistant instead of failing the chat, and stopped commands are cleaned up more reliably.

Assistant

Chat reads large documents during indexing

Chat can now finish reading a large document even while background indexing of that document is still running, instead of timing out before a fresh copy is fetched.

Assistant

Outlook Calendar scheduling works on extension pack deployments

Assistant meeting scheduling, rescheduling, and agenda skills now reliably find the Microsoft Outlook Calendar tools on deployments using the Outlook Calendar extension pack, and handle all-day Outlook events correctly.

Assistant

Manage Zoom meetings through Zoom MCP

The Zoom MCP template now requests meeting create, update, and delete OAuth scopes. After an admin recreates the Zoom MCP server and users reconnect Zoom, authorized users can schedule and manage Zoom meetings through Glean.

Actions

Single approval card for MCP tool calls

Approving an MCP tool call in Assistant or an agent no longer shows the same tool call twice with different icons. Only one card is shown.

Actions

Original source

Release version: 558

This release turns on usage limits for identity provider (IdP) groups by default, adds PowerPoint export for HTML slide decks, brings Claude Sonnet 5.5 to Assistant and Agents, and expands the Model Context Protocol (MCP) server library with 67 new templates. It also delivers new connector betas, stronger permissions enforcement, durable agent-run APIs, and a set of reliability fixes.

Major launches

Set usage limits for identity provider (IdP) groups from the billing dashboard

Admins can now set usage limits for identity provider (IdP) groups directly from the billing dashboard. Many enterprises already manage access through IdP groups, so this brings spend control to the same groups your organization uses everywhere else. Group-mode spend limits are on by default, giving finance and platform owners finer-grained control over consumption without extra setup, and they can still be turned off for your deployment if you prefer a different model.

How to access: Admins set group limits from the billing dashboard in the Admin console. Group mode is on by default and can be turned off per deployment.

Learn more

Admin Capabilities [ROAD-1629]

Export HTML slide decks as PowerPoint presentations

You can now export eligible HTML slide decks as editable PowerPoint (.pptx) presentations directly from the artifact export menu. Instead of rebuilding Glean-generated decks by hand, download a fully editable file and polish it in the tool your team already uses for presentations — speaker-ready formatting included. This closes the loop between generating a deck in Glean and delivering it in your organization's standard format. Access is admin gated, so your admin controls who can use the export.

How to access: Open an eligible HTML slide deck and choose PowerPoint from the artifact export menu. Access is admin gated.

Learn more

Assistant [ROAD-898]

Productboard connector now in public beta

The Productboard connector is now available in public beta. Connect Productboard to Glean to make your product management content searchable alongside the rest of your company knowledge, so product teams can find notes, features, and feedback without switching tools. As a beta release, the connector is still evolving — your feedback helps shape what ships at general availability.

How to access: Admins can add Productboard as a data source from the Admin console.

Connectors

Authorize GitLab with your own account (private beta)

GitLab connector setup now supports user OAuth, so teammates can authorize GitLab with their own accounts and search private groups and projects according to the access they already have. This keeps GitLab results permissions-aware without requiring identity uploads, and it gives each person results that match exactly what they can see in GitLab. The capability is in private beta.

How to access: Private beta — reach out to your Glean account team to participate. Admins then configure user OAuth in the GitLab connector setup.

Learn more

Connectors [ROAD-1658]

Manage your notifications in one place

A new Notifications page in Settings brings your notification preferences together, including email categories and browser notifications. Instead of hunting through Preferences for scattered email options, you get one dedicated place to decide what Glean tells you about and how. This makes it easier to tune notifications to your workflow — keep the alerts that matter and quiet the rest.

How to access: Go to Settings → Notifications in Glean.

Platform [ROAD-1654]

67 new MCP server templates

The Model Context Protocol (MCP) server library keeps growing. This release adds 67 new remote server templates: Acquirepad Commercial Real Estate Activity Intelligence, Highspot, CE Cosmos Deep Dive, SNOMED CT terminology, Orca Security, Jiminny, Fingerprint, Stocklake, Upfluence, Aiven, Windsor.ai for Social Media, Socialinsider, Agility Media Intelligence, Token Terminal, Windsor.ai for Web Analytics, Felt Maps, LunarCrush, Maze, NeetoDesk, QuickFile, FlowGuideAI, Vitally, MyGeotab, Adzviser Google Analytics, Glow for WordPress, Wiv, SaaSify, Sabre Developer Hub, DocketBird, Apaleo, CISO360AI, taskPilot, CometChat, Aphex, CaDoMi.ai, Brainfish, Evarist, eino, GoodDay, Hiveflow, Hgraph, Instrumentl, iSpv, Juicy Profit Analytics, FlowAccount, HelpDocs, Introw Partner Connect, Inbox Monster, iubenda, Fuse Inventory, Jurnie, freee, Horizon, FreightPOP, Lyssna, Hiver Omni, Attribution, Ubersuggest, Cayra, DiligenceSquared, Dataslayer, CreatorDB, Granted, Grid Control, Flockjay, Interfy, and Hardline. Most templates authenticate with per-user OAuth, and several belong to the unverified tier, so admins should review them before enabling them for their organization.

How to access: Admins enable the templates they want from the MCP server library.

Actions [ROAD-1441]

AI & model updates

Claude Sonnet 5.5 in Assistant and Agents

Claude Sonnet 5.5 is now available as a distinct selectable model in Glean Assistant and Agents, giving you another premium Anthropic model option for coding, agentic workflows, research, and long-horizon knowledge work. Sonnet 5.5 does not replace Claude Sonnet 5 and is not part of automatic model routing, so existing model selections stay unchanged. Admins manage model availability and eligible provider routes. This feature may be subject to usage-based pricing.

How to access: Select Claude Sonnet 5.5 from the model picker in Assistant or on supported Agent steps. Admins control model availability in the Admin console.

Learn more

Assistant [ROAD-1754]

Critical updates

Evaluation runs no longer execute write actions

Fixed an issue where internal quality-evaluation runs could execute write actions — for example, sending an email or editing a spreadsheet — through connected third-party tools. Evaluation runs, including voice evaluation runs, now simulate write actions by default and never execute them against connected accounts.

How to access: No action needed. The safeguard applies automatically.

Platform

Externally organized calendar events no longer crawled by default

Outlook Calendar and Teams now crawl organizer-owned calendar events by default, which reduces processing delays for large meetings. Meetings organized outside your crawled user population may no longer be included unless external-event crawling is explicitly enabled for your deployment.

How to access: No action needed for the performance improvement. External-event crawling can still be explicitly enabled for your deployment if you need coverage of externally organized meetings.

Connectors

More accurate Azure Files permissions

Azure Files permission enforcement is now more precise. Glean honors root-scoped and inherited role assignments, matches wildcard permission patterns, and unions compatible share-read and file-read grants across each user or group principal while preserving deny precedence. Glean also honors the storage account's default share-level permission when determining document access, while continuing to enforce Azure deny assignments and New Technology File System (NTFS) permissions.

How to access: No action needed. The improvements apply automatically to Azure Files connectors.

Connectors

Egnyte path restrictions now apply to queued work

Egnyte file tasks now recheck configured path restrictions before processing, so newly excluded files are skipped even if their tasks were queued earlier. This means restriction changes take effect promptly instead of waiting for queued work to drain.

How to access: No action needed. Existing path restrictions are enforced automatically.

Connectors

Enhancements

Control MCP write-action approvals per server

Glean MCP servers gain an enableElicitations setting, on by default. Glean asks the user before write actions using MCP elicitation, since not all hosts provide human-in-the-loop controls. Admins can switch it off for a server to rely on the MCP host's own approval prompts instead. The setting is also available in the Glean MCP server create and update APIs.

How to access: Admins configure the setting per MCP server, or set enableElicitations through the MCP server APIs.

Glean MCP

Durable agent runs through the Agents API

The Agents API now supports durable runs end to end. Runs can continue after the initiating HTTP request disconnects and expose a persisted run ID for polling. The APIs accept tool-approval decisions and resume the same durable run, with idempotent replay of accepted decisions, and support cooperative cancellation of a run — including while it waits for tool approval.

How to access: Available through the Platform Agents API.

Platform

Faster Assistant responses with more precise skill use

Glean Assistant now responds faster and uses skills more precisely, because it no longer loads skill suggestions before every answer.

How to access: No action needed. The improvement applies automatically.

Learn more

Assistant [ROAD-1490]

Advance notice before agent ownership consolidation

Advance notice emails are now on by default for the upcoming single-owner agent migration, so co-owners are informed before ownership is consolidated.

How to access: No action needed. Affected co-owners receive the notice emails automatically.

Agents [ROAD-1558]

Creator logos for open models

Assistant chats now show creator logos for open models, matching the model picker in the composer, so it's easier to see which model produced a response.

How to access: No action needed. Logos appear automatically in chats.

Assistant

Refresh A2A agent cards in place

Admins can refresh an Agent2Agent (A2A) server's agent card directly from the edit drawer, without leaving the form.

How to access: Open the A2A server's edit drawer and select refresh.

Agents

Select all matching departments at once

Admins can select all matching departments in one step when assigning the Departmental Agent Moderator role, instead of picking departments one by one.

How to access: Use the select-all option in the department picker when assigning the role.

Admin Capabilities

Clearer Ironclad connector errors

Ironclad connector validation now reports a clear, actionable IRONCLAD3 error when the admin OAuth token is missing, instead of failing with a generic error. The message tells admins to reconnect Ironclad as an administrator.

How to access: No action needed. The clearer error appears automatically during Ironclad validation.

Learn more

Connectors

Authorize Glean after Slack migrations

Users can now authorize migrated Slack instances after an administrator completes the Slack migration setup, so access recovers smoothly after a workspace migration.

How to access: After your admin completes the Slack migration setup, authorize Slack again when prompted.

Connectors

Find Team chat in Feature rollouts

Admins can now find Team chat in Feature rollouts and follow a link to manage its audience in Assistant settings, making it easier to control who can use the open beta.

How to access: Admin console → Feature rollouts → Team chat, then manage the audience in Assistant settings.

Admin Capabilities [ROAD-1346]

Separate view and edit permissions for model settings

Large language model (LLM) pages in the Admin console now check separate permissions for viewing and updating configuration, so access maps more precisely to each admin role.

How to access: No action needed. The permission checks apply automatically.

Admin Capabilities

More reliable Microsoft 365 crawler authentication

SharePoint and OneDrive crawler operations now use their configured crawler apps, while identity operations continue to use identity credentials, preventing authentication routing issues.

How to access: No action needed. The fix applies automatically.

Connectors

Fresher Intercom indexing

Intercom incremental crawl work is now prioritized over background full-refresh work, improving indexing freshness when both crawl modes overlap.

How to access: No action needed. The prioritization applies automatically.

Connectors

Fixes

Yammer crawls recover

Fixed token refresh for Yammer connectors, which was causing crawls to fail.

Connectors

Date-filtered Outlook searches return results

The MCP outlooksearch tool now documents valid date-filter syntax, so date-filtered email searches return matching results instead of nothing.

Glean MCP

Code Writer on AWS stays available

Internal image cleanup no longer deletes an image still named in a running service's environment, preventing recurring Code Writer failures on AWS.

Platform

People profile photos work in document tools

Glean Document Reader can now open People profile photos hosted on your deployment, so tools such as Create presentation can use them.

Assistant

Reliable Slack workspace metadata crawls

Slack workspace metadata crawls now use a workspace-capable token with the required read permission, including setups that use Gleanbot for private-channel crawling.

Connectors

Original source

Release version: 557

This release makes it faster to find your past chats and artifacts, brings a new browser extension side panel, adds scheduling for manually run agents, and expands the Model Context Protocol (MCP) server library with 25 new templates. It also delivers usage limits for identity provider groups, stronger permissions enforcement across connectors and agents, and a set of reliability fixes.

Major launches

Find past chats and artifacts faster

The redesigned chat search experience organizes everything you've created in Glean into clear categories — Chats, Artifacts, and Meeting Notes — so you can jump straight to the right kind of result. A new sort dropdown lets you order results by relevance or recency, making it easy to pick up where you left off or track down an artifact from weeks ago. Search across your history is now faster and more predictable, whether you remember an exact phrase or only an approximate timeframe.

How to access: Open search from your chat history, then filter by Chats, Artifacts, or Meeting Notes and sort by relevance or recency. No admin action needed.

Learn more

Glean Home [ROAD-1684]

A new native side panel for the browser extension

The Glean browser extension is moving to a native side panel experience that keeps Assistant alongside whatever page you're working on. The side panel also reads page content more accurately — including supported embedded frames and live form and rich-text values — so answers reflect what's actually on your screen. The side panel reaches general availability for standard deployments on October 1, 2026. Deployments using managed rollouts keep admin control — admins can enable it from the Feature rollouts page starting October 1, 2026 — until it reaches general availability for all deployments on October 15, 2026.

How to access: Standard deployments receive the side panel automatically on October 1, 2026. Admins on managed rollouts can enable it from Admin console → Feature rollouts until general availability on October 15, 2026.

Learn more

Glean Extension [ROAD-1693]

Schedule agents configured for manual runs

You can now add schedules to agents configured to run manually. Instead of remembering to trigger a recurring task yourself, set the agent to run on the cadence you need and let it handle the rest — daily digests, weekly reports, or periodic checks all run automatically while the agent keeps its manual-run configuration. This closes the gap between one-off agent runs and fully autonomous agents, giving you a lightweight way to put routine work on autopilot.

How to access: Open an agent set to "Manually run" and add a schedule. No admin action needed.

Agents [ROAD-1709]

Set usage limits for identity provider (IdP) groups

Usage limits at the IdP group level are now available to all Enterprise Flex and Glean Core Suite customers. Many enterprises already use their identity provider as the source of truth for access across tools. Admins can now use those existing IdP groups and group-management workflows to set Glean spend limits for specific teams or groups, giving finance and platform owners finer-grained control over consumption and clearer accountability than a single org-wide cap.

How to access: Admins on Enterprise Flex and Glean Core Suite can configure group limits in Admin console → Usage.

Learn more

Admin Capabilities [ROAD-1629]

Query Databricks with natural language or SQL (Beta)

A new Databricks Genie server template is available in the MCP library, in Beta. Once an admin connects it, Assistant can answer questions about your Databricks data in plain language or run SQL queries directly — bringing warehouse data into the same conversational experience you use for the rest of your company knowledge. Access is admin gated, so your team controls exactly who can query which data.

How to access: Admins connect the Databricks Genie MCP server from the MCP library. Access is admin gated.

Learn more

Connectors & MCP [ROAD-1167]

25 new MCP server templates

The MCP server library keeps growing. This release adds 25 new remote server templates: UserTesting, Kayse, Impala, RankSpot, Loam, exascale.build, Lili, Hologrow, JP Bids, Soracom Knowledge, Paddock Finance, Pagent, RED Atlas, PopHIVE, Tsuga, Speko, Adzviser Matomo, Valency Bond, Docebo, NuMetric, Kernel, Stealthium, Seltz, ProductNow, and RDash. Most templates authenticate with per-user OAuth, and several belong to the unverified tier, so admins should review them before enabling them for their organization.

How to access: Admins enable the templates they want from the MCP server library.

Actions [ROAD-1441]

Model announcement emails for billing admins

Billing admins now receive email announcements when new models become available in Glean, so you can evaluate and plan for new model options without watching release notes or the model picker. Glean sends the announcements automatically, with nothing to configure, helping admins stay ahead of model changes that may affect cost, performance, or governance decisions. Admins can opt out through the "Product news and updates" email preference.

How to access: No action needed. Billing admins receive the announcement emails automatically. Opt out via the "Product news and updates" email preference.

Admin Capabilities [ROAD-1714]

Usage stats on user profiles

Usage stats are now enabled on user profiles, adding activity insights to the profile experience in Glean.

How to access: Open a user profile to see the available stats. No admin action needed.

Glean Home [ROAD-1668]

Assistant suggests your next step

Assistant can now suggest a helpful next step after it responds, so you can keep momentum without figuring out what to ask next. The next-step suggestion engine runs in Glean Key deployments.

How to access: Available automatically in Glean Key deployments. No action needed.

Assistant [ROAD-1722]

AI & model updates

Upgraded model for personalized Cowork suggestions

Personalized Cowork suggestions on Glean Home are now powered by Qwen3.6 for all Glean Key customers, continuing the rollout on AWS. You'll see the improvement automatically — suggestions keep appearing in the same place on Glean Home, backed by the upgraded model.

How to access: Available automatically on Glean Home in Glean Key deployments.

Glean Home [ROAD-1444]

Critical updates

Salesforce redlisted fields enforced across live search and document reads

Redlisted-field enforcement for Salesforce is now on by default for federated (live) search, federated fetch, and rich document reads. If your organization has configured redlisted Salesforce fields, that configuration is now consistently applied across these read paths, so sensitive field values stay excluded everywhere Glean produces results.

How to access: No action needed. Glean enforces your existing redlisted-fields configuration automatically.

Connectors

Agent tool approvals enforced for all tool types

Agent safety checks now honor per-tool approval settings for non-MCP tool packs, such as OneDrive, and MCP tools. Agents that require approval are also correctly greyed out where approval is pending, so approval policies apply uniformly across every kind of tool an agent can use.

How to access: No action needed. Existing per-tool approval settings now apply to all tool packs.

Agents

More accurate Azure Files permissions

Two improvements make Azure Files permission handling more precise. Permission crawling now preserves scoped role-definition IDs, uses stable crawl-node identities, keeps account and share assignments distinct, and re-dirties permissions consistently across pagination. Permission evaluation now unions all role assignments that apply to a user or group, matches action patterns with anchored wildcards, and ensures deny patterns override grants. Together these changes mean search results respect your Azure Files access rules more accurately.

How to access: No action needed. The improvements apply automatically to Azure Files connectors.

Connectors

A more secure file processing sandbox

The file processing sandbox images now run digest-pinned Ubuntu 24.04 with Python 3.12, strengthening the security posture of the environment that handles documents. Document, PDF, optical character recognition, and font processing capabilities are fully preserved.

How to access: No action needed. The hardened images are in use automatically.

Platform

Deprecations & sunsets

Google Sheets write skill turned off

The Google Sheets write skill is now off, after evaluation showed it did not reliably improve output quality. Existing Google Sheets write requests use the remaining supported behavior instead of this skill.

How to access: No action needed.

Assistant

Composio Google Docs copy action turned off

The Composio Google Docs copy action is now off because it exposed no destination parameter, so copies could not reach the intended destination. Use the native Google Drive MCP copy tool instead.

How to access: No action needed. Switch workflows to the native Google Drive MCP copy tool.

Actions

Enhancements

Control who can create and use Skills

Admins get comprehensive governance controls for Skills. A creation selector determines who can create skills — everyone with Skills access, selected users and groups, or admins and skill moderators only. A usage selector controls who can use skills — everyone, selected users and groups, or no one — with existing settings migrated automatically. A single toggle now governs GitHub imports, and the server enforces every audience setting, hiding creation options like file upload, GitHub import, and chat preview from users without permission.

How to access: Admins configure creation and usage audiences in the Skills admin settings.

Admin Capabilities

Rate controls for Git server connectors

The Bitbucket Server, GitLab Server, GitHub Server, and GitHub Enterprise Server connectors now support an aggregate limit on crawl queries per second, letting admins cap Glean's total crawl traffic to protect self-hosted Git servers from load spikes.

How to access: Admins set the aggregate rate limit in each connector's setup form.

Connectors

Keep sensitive query data out

Redlisting for the BigQuery MCP server is now available to all customers, so admins can keep sensitive data out of natural-language and SQL queries on BigQuery.

How to access: Admin gated. Configure redlisting for the BigQuery MCP server.

Learn more

Connectors & MCP [ROAD-1165]

Now on by default

Four previously announced experiences are now on by default for everyone: the updated left navigation, the redesigned model picker, the collapsed Show work summary, and the new homepage.

How to access: No action needed. All four turn on automatically.

Glean Home

Export agent traces to your observability stack

The configuration view for exporting agent traces to external OpenTelemetry Protocol (OTLP) endpoints is now available by default, so admins can send agent traces to their own observability tooling without a support request. Existing permission checks stay the same.

How to access: Admins configure OTLP trace export in the Admin console.

Learn more

Admin Capabilities [ROAD-1559]

Clearer control over agent publishing approvals

The "Publishing an agent requires approval" setting now stays visible whenever default members can create or publish agents, independent of sharing toggles, so admins can always find and manage the approval requirement.

How to access: Admin console → User roles → Default member permissions.

Learn more

Agents [ROAD-1191]

Use IdP groups as audiences for tools and MCP servers

You can now use IdP groups as audiences for Tools and MCP servers, and principals with global Tools write access can browse groups org wide, making it easier to scope tool access to the teams that need it.

How to access: Select IdP groups when configuring tool and MCP server audiences in the Admin console.

Admin Capabilities

Platform API improvements for agent runs

The Platform API adds owner-scoped agent-run retrieval with durable human-in-the-loop flows — start a run, poll for pending approvals, submit decisions, and continue or cancel. Agent Run API responses (public and experimental) now also return citations end to end for both buffered and streamed responses.

How to access: Available through the Platform Agent Run APIs.

Platform

Upload Glean artifacts to OneDrive and SharePoint

The Microsoft 365 upload-file action can now upload Glean-generated PowerPoint (.pptx) and Excel (.xlsx) artifacts to a OneDrive or SharePoint folder you specify, so finished work lands directly where your team keeps it. Uploads to your own OneDrive root continue to work as before.

How to access: Use the upload-file action with a destination folder in agents or Assistant.

Actions

Send Slack DMs and email from Assistant

The Slack direct message and Glean email actions are now enabled on the Assistant surface, so agents running in Assistant can deliver their output to you directly in Slack or by email.

How to access: Use the Slack DM and email actions directly in Assistant.

Actions

Clearer search results for Glean content

Search results now use Library-style gradient icons for artifacts, skills, agents, chats, meeting notes, and projects, and show localized type labels — Document, Interactive, Spreadsheet, Agent, Chat — before metadata, so you can identify result types at a glance.

How to access: No action needed. The updated results appear automatically.

Search

Interface refinements

Paper artifacts are now labeled "Document" instead of "Canvas" to match Library terminology. Additional-tools authorization banners show the connector instance name with a readable fallback. The redesigned composer moves the session-only note into the model tooltip. The Assistant memory attribution indicator now uses a book icon.

How to access: No action needed. The refinements appear automatically.

Assistant

Clearer admin workflows

Billing group-mode switching now shows a deduplicated list of impacted groups and asks for a final confirmation before switching between department and IdP-group usage modes. The inactive-agent bulk-delete report copy now clarifies that it lists agents with zero runs in the selected timeframe and needs manual review. The LLM gateway host selection layout in mobile device management (MDM) settings now matches the latest designs.

How to access: Available automatically in the relevant Admin console and MDM settings.

Admin Capabilities

Validated Google Drive label redlists

Google Drive label redlist setup now verifies that configured choice IDs exist in published label schemas, catching configuration errors before they affect results.

How to access: Validation runs automatically during label redlist setup.

Connectors

Keep embedded chat names in sync

The Web SDK adds a new chat:name_update event, so apps that embed Glean chat can keep custom chat menus in sync as names change.

How to access: Subscribe to chat:name_update in the Web SDK.

Platform

Fixes

More reliable GitHub identity crawls

Identity crawls no longer fail when enterprise-managed-user members have no linked user account, and installation-token validation now uses the same endpoint as crawling, preventing repeated token-refresh fallbacks.

Connectors

Faster file processing under load

Pause activity probes now run off the orchestrator event loop, improving file processing reliability during high load.

Platform

Salesforce crawl resilience

Folder-share not-found errors are no longer retried, so Salesforce crawls don't stall for hours on missing folders.

Connectors

Filter agents by creator

The "By you" and drafts filters now match agents by creator ID or case-insensitive email, fixing empty tabs.

Agents

Audio transcription spend tracking

Audio transcription usage is now recorded in current-month usage totals and counts toward spend limits.

Billing

Original source

Release version: release-2026-09-17

This release adds 182 new vendor Model Context Protocol (MCP) server templates to the MCP server directory, introduces a getting started checklist in the left navigation, moves new MCP setup into the Connectors experience, and brings Monday.com items and updates into Glean search. It also adds Claude Opus 5.5, GPT-6 Sol, and GPT-6 Luna in Assistant and Agents. This release includes agent scheduling and library improvements, new admin controls, security hardening for mobile sign-out and Azure Files permissions, the start of the memory & personalization free trial for Enterprise Flex plans, and the retirement of the legacy v1 browser extension experience.

Major Launches

182 new vendor MCP server templates

The MCP server directory now includes 182 additional vendor Model Context Protocol (MCP) server templates, spanning legal, finance, sales, marketing, real estate, construction, security, developer tools, research, and more. Notable additions include Trello, Stack Overflow, Elicit, TikTok for Business, Alpha Vantage, Whimsical, Interactive Brokers, GrowthBook, Cypress Cloud, DeepJudge, and Paxton Legal Research. Most templates use per-user OAuth, so each person authorizes their own account, and templates that haven't completed Glean verification are visually flagged in the directory so admins can make informed choices.

How to access: Admins explicitly add and enable each server from the MCP server directory in the Admin Console — no servers are turned on automatically. See the full list of supported servers.

Learn more

Glean MCP [ROAD-1441]

Getting started checklist in the left navigation

A new Getting started checklist in the left navigation helps you discover key features and get up to speed with Glean faster. The checklist walks new users through the most useful first steps — like connecting data sources, trying Assistant, and exploring agents — so teams see value from Glean sooner without needing a guided rollout. It's especially helpful for organizations onboarding new employees or expanding Glean to new departments.

How to access: The checklist appears automatically in the left navigation. It's on by default for everyone, and no admin action is required.

Glean Home [ROAD-1705]

MCP setup moves to Connectors

Starting September 22, 2026, all net-new Model Context Protocol (MCP) connections are set up through the Connectors experience instead of the separate Tools area. This brings MCP servers and native connectors into one place, so admins manage all of Glean's data and tool connections from a single, consistent workflow. All customers also see the redesigned connectors table on the Connectors page, which makes it easier to scan and manage existing connections. Existing MCP configurations are unchanged and keep working as before.

How to access: Go to Admin Console → Connectors to set up new MCP connections and see the new connectors table. No migration is required for existing configurations.

Connectors & MCP [ROAD-1706]

Monday.com items and updates in Glean search

Monday.com items, nested items, updates, and replies are now searchable in Glean. Search results open the correct Monday item or board, and Monday permissions are preserved so people only see content they can already access in Monday. This means project details, task conversations, and status updates that previously lived only inside Monday boards are now part of your company-wide search — making it easier to find the full context of a project without switching tools. Content appears after the next successful full content crawl.

How to access: An admin must enable item indexing with crawl.indexItems.enabled=true, and each end user must authorize Monday in Glean under User Settings → Data Sources → Monday to search items and comments.

Learn more

Connectors [ROAD-1711]

Deprecations & Sunsets

Legacy v1 browser extension experience retired

The legacy v1 search-only overlay experience in the Glean browser extension is retired. The option to switch back to the v1 experience has been removed, and individual users still on v1 move to the v2 side panel experience, which offers chat, search, and page context in one place. Deployments still configured for the v1 overlay deployment-wide are not automatically upgraded.

How to access: No action is required for users who already have the browser extension — the v2 side panel experience applies automatically.

Learn more

Browser Extension [ROAD-1643]

Enhancements

Claude Opus 5.5 in Assistant and Agents

Claude Opus 5.5 is now available as a premium Anthropic model in Assistant and Agents. It's designed for coding, agentic workflows, long-running enterprise tasks, research, and knowledge work, and provides an upgrade path from earlier Opus models. Usage may be subject to usage-based pricing.

How to access: Select Claude Opus 5.5 from supported Assistant and agent model settings where premium Anthropic models are available.

Learn more

Assistant & Agents [ROAD-1715]

GPT-6 Sol and GPT-6 Luna in Assistant and Agents

GPT-6 Sol and GPT-6 Luna are now available in Assistant model choice and Agents across Glean Universal Model Key and Customer Key configurations. These models give teams more choice for Assistant conversations and agent workflows. Pricing details are not included in this release note.

How to access: Select GPT-6 Sol or GPT-6 Luna from supported Assistant and agent model settings.

Learn more

Assistant & Agents [ROAD-1724]

Find agents by owner in the Agents library

The Agents library now includes an Owner filter alongside the existing Creator filter. It supports multiple owners — both people and groups — and matches agents owned by any selected owner. This keeps agents discoverable even after ownership transfers, such as when an agent's original creator leaves the company.

How to access: Open the Agents library and use the Owner filter. No admin action is required.

Agents

Scheduled agents keep their subscribers

Agents that run on a schedule no longer pause subscriptions when the agent's owner switches between scheduled and manual runs. Previously, that switch paused every active subscription until each subscriber manually reactivated it. Subscribers now stay subscribed through the change and are still notified of agent changes by email.

How to access: This applies automatically to all scheduled agents. No action is required.

Agents

Simpler agent scheduling

Setting and editing an agent's schedule is now a short guided flow. Setting a schedule asks for the inputs the scheduled run should use and then the cadence; editing one opens on the cadence and confirms the inputs afterwards, so editing a schedule no longer changes what a manual run does. Removing a schedule has moved to an overflow menu next to the schedule action, and steps with nothing to fill in are skipped.

How to access: Open any agent you can schedule and select the schedule action. No admin action is required.

Agents

People profiles stay in embedded chat

Embedded Web SDK chat now keeps people profiles, the org chart, teams, and team pages inside the embedded experience instead of opening separate app.glean.com pages. If your organization embeds Glean chat in an internal portal or app, users can now explore people and team information without leaving that experience.

How to access: This applies automatically to embedded chat built with the Glean Web SDK. No configuration changes are required.

Web SDK

Browser notifications setting has moved

The Allow browser notifications from Assistant and agent runs toggle has been removed from the Chat preferences dialog. Browser notification preferences are now managed in one place.

How to access: Go to Settings → Preferences → Notifications to manage browser notifications.

Assistant

Smoother link typing in the composer

URLs typed in the Assistant composer no longer convert to a link pill before you finish typing, so you can paste or type out a full link without it being cut short.

How to access: This applies automatically in the Assistant composer. No action is required.

Assistant

Browser extension side panel polish

The browser extension side panel now includes an X icon to close the chat bar menu, along with updated placeholder text, for a cleaner and more predictable experience.

How to access: This applies automatically in the browser extension side panel. No action is required.

Browser Extension

Deploy OpenCode through the LLM Gateway with MDM

Administrators can now deploy OpenCode through the LLM Gateway mobile device management (MDM) setup on macOS and Windows, with managed configuration and a selected OAuth client. This lets IT teams roll out OpenCode to managed devices with the same MDM workflow used for other LLM Gateway apps.

How to access: In the Admin Console, open the LLM Gateway MDM setup, select OpenCode, and choose an eligible OAuth client to generate the deployment configuration.

Admin Console

Control over home page background images

Admins can now allow or disallow end-user changes to the home page background image. When custom backgrounds are disallowed, previously selected backgrounds revert to the default, keeping a consistent look across the organization.

How to access: Go to Admin Console → Appearance and set the background image preference for your organization.

Admin Console

SSO credential-expiry alerts on by default

Single sign-on (SSO) credential-expiry email alerts are now enabled by default. Workspace administrators receive advance notice when Azure SSO credentials are nearing expiration, helping prevent authentication disruptions before they affect users.

How to access: Alerts are sent automatically to workspace administrators. No setup is required, and admins can adjust notification preferences if needed.

Admin Console

Clearer usage labels for Memory & personalization

Usage breakdowns in the personal usage view and the billing admin UI now label memory usage as Memory & personalization instead of an internal codename, so the Assistant usage breakdown matches the name used elsewhere in the product.

How to access: The updated label appears automatically in personal usage and billing views. No action is required.

Learn more

Admin Console

Memory & personalization free trial for Enterprise Flex plans

Enterprise Flex plans receive a 30-day free trial of Memory & personalization usage starting September 17, 2026. During the trial, memory usage does not consume FlexCredits. After the trial ends, usage bills against FlexCredits starting October 17, 2026. Billing admins may want to review expected usage before the billing start date.

How to access: The trial applies automatically to Enterprise Flex plans. Review usage in the billing section of the Admin Console.

Learn more

Admin Console

Now on by default for everyone

The following experiences, previously announced on September 15, 2026, are now on by default for all users:

• Updated left navigation, reorganized into tiers with an overflow menu and customizable sidebar
• Redesigned model picker in Assistant, with cost and intelligence ratings and notes
• Collapsed Show work summary in Assistant, which tucks intermediate reasoning and tool calls into a single expandable summary
• New homepage experience, with a dedicated space for company announcements, featured agents, and resources

How to access: These experiences apply automatically. No admin action is required.

Assistant

Security

Mobile sign-out now revokes tokens on the server

Logging out of the Glean iOS and Android apps now revokes the stored refresh token on the server, following the OAuth 2.0 token revocation standard. A signed-out session can no longer be used to obtain new access tokens, strengthening protection if a device is lost or shared.

How to access: This applies automatically when users sign out of the Glean mobile apps. No action is required.

Mobile

Stricter Azure Files permission enforcement

Azure Files now uses bounded keys for Azure role-based access control metadata and requires both Azure role-based access control and file-system permissions to allow access before indexed files are visible in Glean. Because both checks must pass, access may tighten for some users whose visibility previously relied on only one of the two.

How to access: This applies automatically to the Azure Files connector. No admin action is required.

Connectors

Fixes

Inline charts no longer remain blank

Fixed an issue where inline charts could remain blank after a follow-up question until the page was refreshed.

Assistant

GitHub connector sync failures resolved

Fixed GitHub connector sync failures that occurred when a pull request review no longer had an associated GitHub user.

Connectors

Configurable MCP connect timeout

The Model Context Protocol (MCP) connect timeout is now configurable via the mcp.toolsStore.connectTimeoutSecs search config, allowing deployments with slower MCP server initialization to connect successfully.

Glean MCP

Tighter network rules for GCP internal load balancers

Google Cloud Platform (GCP) Capsule internal load balancers are now restricted to private network sources instead of permitting all source ranges at the firewall layer.

Platform

Egnyte transcript content now searchable

Fixed an issue where valid Egnyte Web Video Text Tracks (WebVTT) transcript files could be indexed without searchable transcript content.

Connectors

Original source

Release version: release-2026-09-15

This release adds 284 new vendor MCP server templates, per-user usage details, an updated left navigation, and four new open models, along with new beta API and connector capabilities, enhancements across Assistant, Agents, and the Admin Console, security improvements, and fixes.

Major Launches

284 new vendor MCP server templates

The MCP server directory now includes 284 new vendor templates spanning legal, finance, real estate, HR and recruiting, marketing, healthcare, security, productivity, and developer tools — including Evernote, DeepL, Tableau Cloud, Meltwater, Coursera, DataCamp, GitBook, draw.io, WeTransfer, Reclaim.ai, Neon, WorkOS, AveryGPT, and Legal Brain Agent for researching Japanese laws, precedents, guidelines, and legal books. Unverified templates are visually separated with caution indicators so admins can make informed choices, and the Ticketmaster template ships disabled pending qualification.

How to access: Admins explicitly add and enable each server from the MCP server directory in the Admin Console; most templates use per-user OAuth.

Learn more

Glean MCP [ROAD-1441]

Per-user usage details

Admins can now drill into an individual user's monthly usage, and eligible users can view their own usage trends and breakdowns.

How to access: On by default. Admins drill in from the Usage dashboard in the Admin Console; eligible users view their own usage from Settings.

Learn more

Admin Console [ROAD-1568]

Search and Chat APIs callable without the experimental header (beta)

POST /api/search, GET /api/search/filters, and POST /api/chat no longer require the X-Glean-Include-Experimental header, simplifying integrations that call these endpoints.

How to access: Available in private beta on a phased basis; contact your Glean account team to participate.

Developer Platform [ROAD-1473]

Tableau permissions crawling (beta)

Glean can now crawl Tableau permissions without per-user crawls, keeping Tableau content permissions-aware with less setup.

How to access: Available in private beta and admin-gated; contact your Glean account team to participate.

Connectors [ROAD-1232]

Updated left navigation

Glean's left navigation is reorganized into tiers with an overflow menu. Users can customize their own sidebar, and admins can set the default sidebar layout for their organization from Admin Console → Appearance.

How to access: Managed rollout begins September 15, 2026, with general availability on September 29, 2026. Admins can enable it early from the Feature rollouts page.

Learn more

Search experience [ROAD-1603]

Collapsed "Show work" summary in Assistant

Intermediate reasoning and tool calls now collapse into a single expandable "Show work" summary once a response is ready, keeping conversations easy to read while preserving full visibility into how Assistant reached its answer.

How to access: Managed rollout. Admins can enable it from the Feature rollouts page.

Assistant [ROAD-1679]

Redesigned model picker in Assistant

The model picker now shows cost and intelligence ratings and notes for each model, helping users choose the right model for the task at hand.

How to access: Managed rollout. Admins can enable it from the Feature rollouts page.

Assistant [ROAD-1669]

AI & Model Updates

Four new open models in Assistant and Agents

Four open models are now available: GLM-5.3, a premium open model for complex knowledge work and agentic tasks; GLM-5.3 Flash for faster, cost-efficient everyday workflows; Kimi K3 for artifact creation and design-heavy workflows; and DeepSeek V4.1 Flash for fast, highly cost-efficient general knowledge work and agentic use. GLM-5.3 and Kimi K3 are premium models; GLM-5.3 Flash and DeepSeek V4.1 Flash are standard models. GLM-5.3 Flash, DeepSeek V4.1 Flash, and Kimi K3 are the first open models with primary placement in the model picker. All four are served through Glean's US-based inference-hosting providers and follow the regional controls for open models (see manage open models by region).

How to access: For Glean Universal Model Key deployments, go to Admin Console → Models → Other open models, turn on open models trained by Chinese companies, and select Save. Access can also be restricted by department or team, or blocked separately for Assistant or Agents.

Learn more

Assistant [ROAD-1670]

Enhancements

Document type filter grouping

The Type filter now groups Glean documents with other document types under a single Document option, making it easier to narrow results.

How to access: Use the Type filter on the search results page.

Search

Per-application usage breakdown in billing drill-down

Admins can now break down a user's model usage by application in the billing drill-down, with mixed denominations separated automatically.

How to access: Open a user's usage drill-down from the Usage dashboard in the Admin Console.

Admin Console

Unsubscribe from agent notification emails

Users can now unsubscribe from agent notification emails, including agent updated or paused notices, approval digests, admin-disabled notices, bulk-delete completion emails, and rate-limit emails.

How to access: Use the unsubscribe option in any agent notification email.

Agents

GitHub skill import guidance

The skill import flow now explains how to pick a specific skill folder when a pasted GitHub URL points at a parent directory.

How to access: Go to Settings → Skills and select Import from GitHub.

Learn more

Assistant [ROAD-1407]

Meeting notes default sorting by created date

Meeting notes now sort by created date by default, so the most relevant notes are easier to find.

How to access: Open Meeting notes; the new default sorting applies automatically.

Assistant

Clearer voice-chat icon in recent chats

A waveform icon now identifies voice chats in your recent chats, making them easier to spot at a glance.

How to access: Applies automatically in your recent chats list.

Assistant

Agent schedules stay visible after switch to manual run

When an agent is switched to manual runs, activated users can still see, pause, or resume its schedule.

How to access: Open the agent's schedule view.

Agents

Admin Console polish

The Admin Console now has consistent page headers and spacing, icons on connector action buttons, and improved Add connector dialog sizing.

How to access: Applies automatically across the Admin Console.

Admin Console

OneDrive artifact export

OneDrive upload actions now export supported PowerPoint and Excel artifacts to the user's OneDrive root, rename conflicting files automatically, and handle large files with retries and parallel uploads.

How to access: Use a OneDrive upload action on a supported artifact.

Tools

Clearer MCP write-action approval prompts

Approval prompts for MCP write actions now name the action by its human-readable title and the connected app instance, so users know exactly what they're approving.

How to access: Applies automatically when an MCP write action requests approval.

Glean MCP

Artifacts and image generation tools on by default for Glean MCP

The artifacts and image generation tool bundles are now enabled by default on the default Glean MCP server and on new custom servers.

How to access: On by default; admins can turn either bundle off in MCP server settings.

Glean MCP

Web SDK chat bar capabilities

The new customizations.features.chatBarCapabilities option lets you choose which chat bar controls the embedded chat widget shows. Omitting the option leaves existing behavior unchanged.

How to access: Set the option in your Web SDK configuration.

Developer Platform

Zendesk connector save-then-authorize setup

Zendesk connector setup now uses a save-then-authorize flow for a smoother configuration experience.

How to access: Set up or reconfigure the Zendesk connector in the Admin Console.

Connectors

Ironclad setup guidance for na1 customers

Ironclad setup guidance now clarifies that na1 customers should leave the Server domain field empty unless OAuth requires it.

How to access: See the guidance in the Ironclad connector setup flow.

Connectors

Per-host LLM Gateway selection for managed deployments

Generated macOS and Windows mobile device management (MDM) scripts now configure only the LLM Gateway hosts you select.

How to access: Select hosts when generating MDM scripts in the Admin Console.

Admin Console

Security

Zendesk OAuth bounded token lifetimes

Zendesk OAuth now uses bounded token lifetimes with refresh-token renewal, reducing the exposure window of long-lived credentials.

How to access: Applies automatically to Zendesk OAuth connections.

Connectors

Glean MCP agent tools respect admin approval policies

Glean MCP agent tools now respect admin approval policies for write actions, so administrative controls apply consistently.

How to access: Applies automatically based on your configured approval policies.

Glean MCP

Azure Files bounded access keys

Azure Files access now uses bounded keys for Azure role-based access control metadata, and users must be allowed by both Azure role-based access control and file-level permissions to see content.

How to access: Applies automatically to the Azure Files connector.

Connectors

GCP dedicated Cloud Build IAM roles for Secure Web Proxy forwarding rules

Dedicated Cloud Build IAM roles on GCP can now manage the Service Directory records created for Secure Web Proxy Private Service Connect forwarding rules, keeping build permissions narrowly scoped.

How to access: Applies automatically to GCP deployments.

Platform

Fixes

Salesforce nested public-group permissions

Fixed an issue where nested public-group hierarchies in Salesforce could over-grant permissions.

Connectors

Slack Enterprise Grid channel permissions

Slack Enterprise Grid now uses the real-time Slack identity source as the authority for channel permissions.

Connectors

Collections shared with everyone appear in search

Collections shared as "Everyone can edit" now appear in search for non-owners.

Search

Glean MCP search on search-only deployments

Fixed Glean MCP search and document retrieval for search-only deployments; MCP Chat is now offered only when Chat is enabled.

Glean MCP

LLM-only mode no longer uses read tools

LLM-only (no sources) mode no longer uses the document reader or MCP read tools, so responses stay strictly source-free as expected.

Assistant

Original source

Release version: release-2026-09-10

This release adds 64 new vendor MCP server templates to the MCP server directory, turns on audio artifacts by default for eligible deployments, brings the Coda (Service Account) MCP template to public beta, and delivers enhancements and fixes across the Admin Console, connectors, platform, and developer surfaces.

Major Launches

Connect 64 new vendor tools through the MCP server directory

The MCP server directory now includes 64 new vendor MCP server templates, so your teams can bring trusted third-party data and tools directly into Glean conversations. New options span sales, finance, legal, healthcare, security, engineering, and marketing vendors, including Salesloft, CircleCI, MongoDB Atlas, BigID, Fitch Solutions, S&P Global Adaptive Retrieval, Oxford Economics, D&B Finance Analytics, LawVu, Canny, Funnel, Boomerang, AlphaSights, Staircase AI by Gainsight, and many more. Most templates connect through per-user OAuth, so each person accesses only the vendor data they are already permitted to see. For setup instructions, see Connect remote MCP servers to Glean.

How to access: Admins can add these servers from the MCP server directory in the Admin Console. Most templates use per-user OAuth sign-in.

Learn more

Glean MCP ROAD-1441

Audio artifacts enabled by default

Audio artifacts are now enabled by default for eligible Glean Key deployments. You can turn chats, documents, and research into playable audio briefings and podcasts without waiting for your admin to switch the feature on, making it easier to catch up on work while commuting or between meetings. Explicit workspace access restrictions are still honored, so organizations that have limited who can use artifacts keep their existing controls.

How to access: On by default for eligible Glean Key deployments; ask Glean to create a podcast or audio version of your content. Admins can opt out in the Admin Console.

Learn more

Assistant ROAD-1373

Coda (Service Account) MCP template in public beta

The Coda (Service Account) MCP server template is now in public beta. Admins can add it for service-identity agents without requiring an allowlist, letting agents act with their own scoped Coda credentials instead of borrowing an individual user's access. This gives teams a governed way to build agents that read and update Coda content as part of automated workflows, while keeping credentials centrally managed.

How to access: Admins can add the Coda (Service Account) credential template from the Admin Console; no allowlist is required.

Learn more

Agents ROAD-1542

Critical Updates

Memory for more personalized responses

Glean now uses memory to provide more relevant, personalized responses over time. Activity-based memory uses permitted workplace activity to personalize responses in the background and consumes FlexCredits for Glean Core Suite (GCS) plans on supported models. Saved memories and preferences are explicitly provided by users, included with every seat, and do not consume FlexCredits. For Enterprise Flex plans, activity-based memory is free through October 16, 2026, and begins consuming FlexCredits on October 17, 2026. The same experience is planned for self-hosted deployments by the end of September 2026.

How to access: No action is required. Admins can manage activity-based memory under Advanced Features and review plan and usage reporting before the Enterprise Flex billing date. Additional controls to manage or disable activity-based memory will be available soon.

Assistant Memory

Deprecations & Sunsets

Deep Research follow-up query suggestions removed

Follow-up query suggestions no longer appear after Deep Research responses in Chat.

How to access: No action is required; the suggestions no longer appear.

Assistant

Bulk employee indexing endpoint removal extended to January 15, 2027

The removal date for the POST /api/index/v1/bulkindexemployees endpoint has been extended from October 15, 2026 to January 15, 2027, giving integrators more time to migrate.

How to access: No immediate action is required; migrate any integrations off this endpoint before January 15, 2027.

Developer Platform

Agent sharing defaults return to legacy audience permissions

Agent sharing now uses legacy audience permissions by default. Granular sharing controls remain in beta, and workspaces that explicitly enabled them keep their configured controls.

How to access: No action is required. Workspaces with granular sharing controls enabled are unaffected.

Agents ROAD-1557

Enhancements

Clearer usage-limit alert thresholds

Usage-limit controls in the Admin Console now make selected alert thresholds clearer, so billing admins can see at a glance which thresholds will trigger notifications.

How to access: Available automatically in the Admin Console usage-limit settings.

Admin Console

Coda Docs connector renamed to Superhuman Docs

The Coda Docs connector is now named Superhuman Docs throughout the connector catalog, admin setup experience, and related UI surfaces, reflecting the vendor's product rename. The underlying connector key, API terminology, and setup field requirements are unchanged, so existing configurations continue to work without any changes.

How to access: No action is required; the new name appears automatically in the connector catalog and connector setup.

Connectors

Profile custom fields displayed for all users

Custom fields configured on people profiles now display below the About section for all users, making it easier to find team-specific profile details.

How to access: No action is required; custom fields appear automatically on people profiles below the About section.

Search

Add artifacts to projects from the Library

You can now add artifacts to projects directly from the Library, where they appear under Sources. The Artifacts tab in projects is now named Outputs, making it clearer which items are project inputs and which are generated results.

How to access: Open the Library, select an artifact, and add it to a project. It appears under the project's Sources.

Assistant

O365 people-profile URL templates support metadata field paths

The O365 connector's userProfileBaseUrl setting now supports arbitrary metadata field paths in the form %{field.path}, so admins can build people-profile links from nested metadata fields rather than a single substitution value.

How to access: Configure userProfileBaseUrl with %{field.path} placeholders in the O365 connector setup.

Connectors

Web SDK controls for agents, projects, and skills in the embedded widget

The Glean Web SDK now supports turning off agents, projects, and skills in the embedded chat widget through customizations.features. Unlike existing options that hide individual UI elements, these remove the feature as a whole concept, including navigation, chat bar controls, and direct routes. All three default to enabled, so existing embeds are unaffected.

How to access: Set the corresponding flags in customizations.features in your Web SDK configuration.

Developer Platform

Private Service Connect support for Secure Web Proxy integrations on GCP

GCP deployments using dedicated custom IAM roles can now provision and manage Private Service Connect forwarding rules for Secure Web Proxy integrations, expanding supported network configurations for security-conscious deployments.

How to access: Available for GCP deployments using dedicated custom IAM roles; no additional configuration is needed beyond your existing deployment setup.

Platform

Audience selectors included in action product-snapshot log events

Action product-snapshot log events now include the user, department, and external-group audience selectors configured for Chat, Agents, and Platform, giving audit teams a fuller picture of how action visibility is configured. This is a logging-only change and does not alter action visibility or runtime access control.

How to access: Included automatically in exported product-snapshot log events.

Platform

Assistant activity events scrubbed in customer event logs

Assistant activity events (GLEANBOTACTIVITY) exported to customer event logs are now scrubbed when raw event logging is disabled for the deployment. Previously these events were always emitted in raw form, so user email addresses and department names could appear in logs that were expected to contain only obfuscated identifiers.

How to access: Applied automatically when raw event logging is disabled for your deployment.

Platform

Generated answer links open the correct file

Generated answer links to downloadable files and artifacts now preserve the file or artifact link instead of opening an unrelated Knowledge Graph entity panel.

How to access: No action is required; the corrected link behavior applies automatically.

Search

Fixes

Chat-history privacy opt-outs apply across all identities

Fixed an issue where users with multiple email identities could not re-enable chat history. Privacy settings now apply across all of a user's identities, and the settings page reflects an opt-out from any identity.

Assistant

Project access preserved after identity changes

Project permission checks now match alias-backed user roles against both stored and current canonical identities, preventing authorized owners, editors, and viewers from losing access after an identity change.

Assistant

Agent share lists no longer show duplicates

Agent share access lists now use consistent person identifiers, preventing duplicate entries when searching for teammates. Sharing and unsharing agents also works correctly when legacy stored identifiers differ from the canonical person ID.

Agents

Slack private channels discovered in custom no-bot setups

Fixed an issue that prevented Slack private channels invited to the Glean bot from being discovered in custom no-bot Slack setups.

Connectors

LumApps deep links resolve to the indexed document

LumApps deep links pasted from the browser now resolve to the indexed document in Glean Assistant, Agents, and Doc Reader.

Connectors

Original source

This release adds 31 new vendor MCP server templates to the Glean MCP directory, brings GPT-6 Astra to Assistant and Agents, and introduces single-owner agent governance — alongside enhancements across connectors, the browser extension, and admin tooling.

Major Launches

Connect 31 new vendor MCP servers from the directory

The Glean MCP server directory now includes 31 new vendor templates, so your teams can bring more third-party tools directly into Assistant and Agents. New additions include ClickHouse Cloud, PineGap, OneSignal, Demandbase, ServiceNow (customer-defined MCP Server Console endpoints), Coevera, Nutshell CRM, Adobe Customer Journey Analytics, Tabs, MuleSoft Platform, Onplana, Commercient Data Lake and Data Hub, Pensieve, Aleph, AI for Work, Adobe Workfront, Windsor.ai for CRM, Reevo, CredCore Tusk Liquid, Revealed, Matia, Suger, Smarter Drafter, Bloomberg Law, Elite Site Optimizer, Parallel Search, Zylo, Voicenter, Orca Scan, Fullstory, and Parallel. Most templates use per-user OAuth; Parallel Search connects anonymously, and some vendors (Demandbase, Bloomberg Law, Adobe Workfront) require vendor-side enablement or entitlement.

How to access: In the Admin Console, open the MCP server directory, select a vendor template, and connect and enable it for your organization.

Learn more

Glean MCP [ROAD-1441]

Choose the right API-key header when configuring MCP servers

When you configure an MCP server that authenticates with an API key, you can now choose which header Glean uses to send the credential — for example Authorization or X-API-Key — from a dropdown in the configuration flow. This removes a common source of setup ambiguity and failed connections when a vendor's server expects the key in a non-default header. The option is available to admins by default, so you can match each server's expected header format without workarounds or support tickets.

How to access: In the Admin Console, open your MCP server's API key configuration and select the header format from the dropdown.

Learn more

Glean MCP [ROAD-1640]

Configure default sidebar navigation in the new Appearance panel

A new Appearance panel in the Admin Console lets you configure which navigation items appear by default in the upcoming customizable sidebar. You can decide ahead of time which destinations your users see first — so when the new left navigation reaches your organization, it reflects the workflows that matter most to your teams rather than a generic default. The panel is available to admins now, giving you time to review and set your preferred defaults.

How to access: Go to Admin Console → Appearance → sidebar navigation defaults.

Admin Console [ROAD-1603]

AI & Model Updates

GPT-6 Astra is available in Assistant and Agents

You can now use GPT-6 Astra in Glean Assistant and Agents. Astra is strongest on long-running artifact workflows and complex reasoning, coding, and agentic work, and delivers a quality advantage over GPT-5.6 Sol and Terra, with tradeoffs in latency and input-token cost. GPT-6 Astra is generally available for Enterprise Flex, Glean Core Suite, and Enterprise (Customer Key) customers; it is not available on Enterprise (Glean Key) or Enterprise Flex in a preview period, and Azure support is coming shortly after.

How to access: Admins can enable GPT-6 Astra for eligible users; once enabled, users can select it in Assistant and Agents.

Learn more

AI Models [ROAD-1685]

Critical Updates

Agents now have a single owner

Agents now enforce a single owner per agent by default, giving each agent one clear point of accountability. Adding multiple owners is no longer supported; instead, owners can transfer ownership through the share dialog, and the previous owner automatically becomes an Editor with full edit access. Existing agents with multiple owners have a two-week grace period, after which they are automatically migrated to a single owner — the agent's creator when possible, otherwise the first owner alphabetically. The original creator is preserved as audit metadata in all cases.

How to access: No action is needed — this change applies automatically, and existing multi-owner agents migrate after the grace period.

Agents [ROAD-1558]

Enhancements

Review changes before publishing an agent

The agent builder's publish modal now shows you the changes that will be published, so you can review exactly what's changing before it goes live. This diff view helps you catch unintended edits, confirm the scope of an update, and publish with confidence — especially useful for agents shared broadly across your organization.

How to access: No action is needed — the diff view appears automatically in the publish modal when you publish an agent.

Agent Builder

Export meeting notes from every tab in the Library

When you open a meeting note artifact from the Library that doesn't have an associated chat, export actions now appear on every tab — matching the experience on the standalone summary page. You no longer need to navigate to a specific tab or reopen the note elsewhere to export your transcript, summary, or notes.

How to access: Open a meeting note from the Library; export actions are available on every tab.

Learn more

Assistant [ROAD-769]

Download the Glean MCP installer from Connected Apps

You can now download the Glean MCP installer directly from Connected Apps to configure Glean MCP on supported macOS and Windows devices. The one-click installer download is available from the MCP connect card and is on by default, so you can set up Glean MCP on your device without waiting for a centrally managed deployment.

How to access: Open Connected Apps and use the download option on the MCP connect card.

Learn more

Glean MCP

Service Cloud migration banner shows its full content

The Service Cloud migration banner in the browser extension now expands vertically to show its full content, instead of hiding part of the message behind an internal scrollbar. The banner keeps its existing width and layout, so you can read the complete migration guidance at a glance.

How to access: No action is needed — the banner displays its full content automatically in the browser extension.

Learn more

Browser Extension [ROAD-1461]

Pasted links appear as compact pills in the composer

When you paste a URL into the Assistant composer that doesn't resolve to a Glean document, it now appears as a compact link pill with smart truncation, keeping long URLs from cluttering your message. Hover over the pill to see the full URL, and rest assured the original destination is preserved exactly when you submit your message.

How to access: No action is needed — paste any URL into the Assistant composer to see the new link pills.

Chat

Ask table-level questions about entire SharePoint Lists

Assistant and Agents can now read a SharePoint List as a single aggregate table document. Paste a list's landing-page URL and Glean reads the entire list as one table — removing the previous cap of roughly 200 items — so you can ask table-level questions like totals, filters, and cross-row comparisons over the full list.

How to access: Requires SharePoint Extension Actions setup; once configured, reference a SharePoint List by its landing-page URL in Assistant or Agents.

Connectors [ROAD-1662]

Azure Files search supports custom metadata and ownership facets

If your organization uses the Azure Files connector, search now supports custom metadata fields (x-ms-meta-*) as searchable custom fields. In addition, author and owner facets are derived from resolved NTFS file ownership when user identity metadata is available, so you can filter Azure Files results by who owns them.

How to access: No action is needed — custom metadata fields and ownership facets are picked up automatically where the metadata is available.

Connectors

Configured-agent status in the extension admin console is now reliable

The browser-extension admin console now keeps configured-agent status and counts accurate. Agents that have been deleted, are inaccessible, or can't be retrieved are clearly labeled with role-appropriate recovery guidance, so you know exactly what to fix and how. New configurations also exclude legacy static workflows and show only configured agents, giving you a clean, trustworthy view of what your users will see in the extension.

How to access: No action is needed — the improvements appear automatically in the browser-extension admin console.

Learn more

Browser Extension [ROAD-1461]

Generate MDM deployment scripts from a central page

Workspace admins can now use a central MDM deployment page in the Admin Console to generate macOS and Windows deployment scripts for enabled MCP Servers and LLM Gateway. Instead of assembling deployment configurations by hand for each platform, you generate ready-to-use scripts in one place and push them through your MDM provider — making managed rollout of Glean MCP across your device fleet faster and less error-prone.

How to access: Open the MDM deployment page in the Admin Console and generate scripts for your enabled MCP Servers and LLM Gateway.

Learn more

Admin Console

Expiring Google Drive permissions stay accurate through large crawls

Google Drive permission crawls now correctly handle paginated, time-limited access grants for files and shared drives. When a crawl spans multiple pages of results, expiring permissions are tracked accurately all the way through, so search results continue to respect exactly who has access — even in large Drive environments with many temporary access grants.

How to access: No action is needed — this improvement applies automatically to Google Drive permission crawls.

Connectors

Agent sharing is on by default for new customers

For new Glean customers, agent sharing is now enabled by default for all users unless the admin explicitly scopes it. This means teams at newly onboarded organizations can share the agents they build right away, without waiting for an admin to switch sharing on. Admins retain full control and can scope or restrict sharing at any time from agent access settings.

How to access: No action is needed for new customers — sharing is on by default; admins can adjust scope in agent access settings.

Learn more

Agents [ROAD-1557]

Fixes

Approval-gated agent actions report their true status

Agent actions that require user approval no longer report success while still awaiting approval; deferred actions now correctly show as pending approval.

Agents

MCP SSE tool discovery no longer drops sessions

MCP SSE tool discovery now keeps the session alive through initialization and tool listing, preventing session-not-found failures during tool refresh.

Glean MCP

PagerDuty incident crawls are faster

Fixed PagerDuty connector pagination that caused excessive API calls and slow incident crawling.

Connectors

NetSuite crawls put less load on your source

NetSuite full crawls now run daily by default while hourly incremental crawls preserve freshness, reducing load on the NetSuite API.

Connectors

Chat history can be re-enabled across multiple identities

Users with multiple email identities can now re-enable chat history; privacy settings apply consistently across all of a user's identities.

Chat

Original source

Major Launches

Bitbucket Code Writer is now generally available

Code Writer support for Bitbucket is now generally available for eligible Glean Key deployments. Code Writer lets Glean make AI-powered code changes directly in your Bitbucket repositories, so engineering teams can turn requests into reviewable code changes without leaving the tools they already use. With this launch, teams working in Bitbucket get the same AI-assisted development workflow previously available for other source-control platforms, helping developers move from an idea to a proposed change faster while keeping review and approval in your existing repository workflow.

How to access: Admins must configure Bitbucket for Code Writer by following the Code Writer setup instructions. Available for eligible Glean Key deployments.

Learn more

Code Writer [ROAD-1589]

Adobe Workfront connector available in public beta

You can now connect Adobe Workfront to Glean. The new connector, available in public beta, indexes Workfront projects, tasks, issues, documents, and updates, making your work-management content searchable alongside the rest of your company knowledge. Glean respects each user's Workfront access permissions, so people only see the content they are already permitted to view in Workfront. This helps teams find project status, task details, and discussion updates without switching tools or digging through Workfront manually.

How to access: In the Admin Console, go to Add data source and select Adobe Workfront, then configure the connector with OAuth2 credentials. Admin configuration is required.

Learn more

Connectors [ROAD-1553]

AI & Model Updates

Claude Fable 5.1 available for Fable 5 customers

Claude Fable 5.1 is now available in Assistant and Agents for customers already enabled for Fable 5. Fable 5.1 takes the primary position in the model picker, while Fable 5 remains available under More models, so you can keep using the previous model if you prefer. This gives eligible teams access to the newest Fable model generation for everyday assistant conversations and agent runs without any configuration changes.

How to access: No admin action is needed. Customers enabled for Fable 5 can select Claude Fable 5.1 from the model picker in Assistant and Agents.

Learn more

AI Models [ROAD-1672]

Gemini 3.8 Flash available in Assistant and Agents

Gemini 3.8 Flash is now available in Assistant and Agents for both Glean Key and Customer Key deployments. Gemini 3.8 Flash is a fast, premium model option, giving you quicker responses for everyday questions and agent runs while keeping access to higher-capability models for more complex work.

How to access: No admin action is needed. Users in eligible deployments can select Gemini 3.8 Flash from the model picker in Assistant and Agents.

Learn more

AI Models [ROAD-1676]

Critical Updates

Update generated Platform Chat SDK integrations

Generated Platform Chat SDKs now separate non-streaming and streaming operations and expose typed server-sent events without buffering the response body. This is a breaking change for generated clients: developers should regenerate their SDK clients and update streaming integrations to use the separated operations and typed event handling.

How to access: SDK developers should regenerate Platform Chat SDK clients and update code that calls chat operations or consumes streaming responses.

Learn more

Developer Platform

Deprecations & Sunsets

Client API agents endpoints deprecated

The agents search, get, schemas, and run endpoints under /rest/api/v1/agents are deprecated. Use the Platform Agents API at /api/agents instead; the create, edit, and import endpoints remain supported. Removal of the deprecated endpoints is scheduled for April 15, 2027, so integrators have time to migrate.

How to access: API integrators should migrate affected integrations to the Platform Agents API at /api/agents before April 15, 2027.

Learn more

Developer Platform

Enhancements

Search API errors clarify pagination limits

Search API errors now distinguish between an invalidly large page size and pagination that is too deep, making invalid pagination requests easier to diagnose and correct.

How to access: No action is needed; the clearer error responses apply automatically to Search API requests.

Developer Platform

Issue titles have more room in the Issues table

The Issues table now gives issue titles more horizontal space while remaining usable across supported screen sizes, making issues easier to scan.

How to access: No action is needed; the improved layout appears automatically in the Issues table.

Admin Console

Scheduled agent subscriptions stay active through edits

Scheduled agents no longer pause active subscriptions when their owner edits input fields or switches between chat and form triggers. Subscribers are still notified of changes and asked to confirm saved inputs when they resume the agent.

How to access: No action is needed; subscriptions remain active automatically when owners edit agent inputs.

Agents

View usage for published agents before setting a limit

The View usage action is now available for published agents with the feature enabled and the required edit permissions, even before a usage limit is configured.

How to access: Open a published agent and select View usage; requires the feature to be enabled and edit permissions on the agent.

Agents

Platform API examples use named placeholders

Platform API reference examples now show named placeholders, such as {agent_id}, for caller-supplied values, making it clear which request fields must be replaced with values from your own instance.

How to access: No action is needed; the updated examples appear in the Platform API reference documentation.

Developer Platform

People suggestions reflect organizational proximity

Personalized people ranking now considers organizational-chart distance, improving the relevance of people suggestions in autocomplete.

How to access: No action is needed; the improvement applies automatically to people suggestions in autocomplete.

Search

Host app usage links open filtered MCP Insights

Host app usage links now open MCP Insights at the usage breakdown with the selected host application filter already applied, so you land directly on the relevant view.

How to access: Select a host app usage link; MCP Insights opens with the host application filter already applied.

Glean MCP

Glean Protect navigation scales beyond the Admin Console

Glean Protect now has a scalable navigation area outside the Admin Console for SCMs and super admins, with sectioned menus that make security capabilities easier to find as they grow.

How to access: SCMs and super admins can use the new Glean Protect navigation area; sections appear as they are enabled for your deployment.

Glean Protect

Claude usage labeled correctly in Host apps

The Host apps tab now correctly recognizes Claude for Teams and Enterprise usage as Claude, so active Claude usage no longer appears as Connect.

How to access: No action is needed; the corrected labeling appears automatically in the Host apps tab.

Admin Console

GitLab token check labeled as format validation

GitLab connector setup now labels the access-token check as format validation, clarifying that a passing check confirms the token format but does not authenticate the token.

How to access: No action is needed; the clarified label appears in GitLab connector setup.

Connectors

MCP Insights estimates token and cost savings

Insights → MCP Servers now shows estimated token and cost savings from Glean Search for deployments with at least 1,000 weekly Glean Search MCP queries, helping you quantify the value of Glean Search in MCP workflows.

How to access: Open Insights → MCP Servers; estimates appear for deployments with at least 1,000 weekly Glean Search MCP queries.

Glean MCP

Index code files that have no extension in Git connectors

Git setup pages now let admins list code filenames that have no extension, such as Dockerfile and Makefile, for body indexing, so these common files become searchable.

How to access: In Git connector setup, list the filenames without extensions that you want included for body indexing.

Connectors

Clearer Marketo connector setup instructions

Marketo connector setup instructions now identify the admin entry point, explain the least-privilege permission hierarchy, and show how to generate and retrieve the Client ID and Client Secret.

How to access: Follow the updated instructions in Marketo connector setup.

Connectors

Marketo setup requires completed authorization

Marketo setup now remains incomplete after credentials and the roles workbook are saved until the administrator successfully completes authorization, preventing an unverified setup from appearing complete.

How to access: Complete the authorization step in Marketo connector setup to finish configuration.

Connectors

Configure Box Inclusion Rules in Manage Data

Box administrators can now configure Inclusion Rules directly in Manage Data. Full crawls can start from configured folders and reuse discovered folder owners for more efficient nested-folder crawling.

How to access: In the Admin Console, open Manage Data for Box to configure Inclusion Rules.

Connectors

Fixes

Data source token refresh capacity is better isolated

Data source instance token refreshes now use their dedicated queue range instead of sharing the action-pack queue range, improving refresh capacity isolation.

Platform

Custom action OAuth tokens renew more reliably

Custom actions using OAuth Admin with the client-credentials grant can now renew expired access tokens even when an existing token record has no refresh token, preventing repeated prompts to authenticate again and permanent unauthorized states.

Actions

Agent Insights validates time-saved feedback

Agent Insights now rejects time-saved feedback that is zero, negative, non-numeric, or over seven days, preventing invalid submissions from producing extremely large metrics.

Insights

Agent Insights excludes extreme time-saved outliers

Agent Insights now excludes time-saved outliers over seven days from precomputed totals, keeping time-saved metrics accurate for reporting.

Insights

Typeform MCP uses OAuth sign-in

The Typeform MCP server now uses OAuth sign-in with Dynamic Client Registration. Production deployments must allowlist Glean's OAuth callback domain.

Glean MCP

Oversized click-boost entries no longer block indexing

Glean now limits click-boost metadata payloads and drops only oversized click-boost entries, preventing large payloads from blocking otherwise valid documents from indexing.

Search

Deletion tasks handle non-ASCII document IDs

Asynchronous deletion tasks now preserve non-ASCII document IDs, and URL-data cleanup skips IDs that exceed the database column limit instead of failing the cleanup operation.

Platform

Auto Mode supports multi-word MCP tool names

Auto Mode agents can now invoke selected multi-word MCP tools when the tool name is already formatted with spaces.

Agents

SharePoint agent knowledge sources appear correctly

SharePoint sites added as agent knowledge sources now appear as valid entries in the agent builder instead of being incorrectly marked invalid.

Agent Builder

Smoother caret navigation around pasted links

Chat composer caret navigation around pasted document links now works naturally, allowing you to place the cursor before or after a document link and continue editing.

Chat

New Google Calendar connectors pass credential checks

New Glean-hosted Google Cloud projects now enable the Google Calendar API during setup, preventing credentials health-check failures for newly configured Google Calendar connectors.

Connectors

More Confluence link types recognized as readable

Confluence folder, database, and whiteboard links that include a page-title slug are now recognized as readable document links.

Connectors

Veeva Vault ownership resolved more accurately

Veeva Vault document ownership now recognizes both supported owner roles, so documents use the Vault owner instead of the version author when that owner is available.

Connectors

Custom providers support more model pairings

Custom-provider configurations now support additional GPT, Claude, and Gemini large/small model pair combinations, including GPT-5.6 Luna, Claude Sonnet 4.6 with Haiku 4.5, and Gemini 3.1 Pro with Gemini 3.6 Flash.

AI Models

GCP Spark batch jobs start successfully

GCP Spark batch jobs using Dataproc Serverless 3.0 now have the required runtime IAM permissions and can start successfully.

Platform

Original source

This release lets you build Glean agents directly from coding tools like Cursor and Codex, brings the Fathom connector to public beta, introduces Manager Insights, updates Gemini and GPT model options, and delivers a broad set of enhancements and fixes across chat, connectors, security, and administration.

Major Launches

Build and run Glean agents from your coding tools

You can now build, test, and run Glean agents directly from coding assistants such as Cursor, Claude, or Codex, without leaving your development environment. This lets developers create agents that draw on company knowledge using the tools they already work in every day, while the Agents library provides step-by-step guidance for creating agents this way. The result is a faster path from an idea to a working agent for technical teams, with the same agents available afterward in Glean for the rest of the organization to use and share.

How to access: Open the Agents library in Glean to find agent-creation guidance, then follow the headless builder documentation to create and run agents from Cursor, Claude, or Codex. Available automatically to users who can build agents.

Learn more

Agents [ROAD-1451]

Edit, save, and download embedded skills in Auto Mode Agent Builder

Auto Mode Agent Builder now gives you direct control over the skills embedded in your agents. You can open an embedded skill, review and edit its instructions, save your changes, and download the updated skill file. Downloaded skills can be reused in other agents, shared with teammates, or checked into version control so your team can track how agent behavior evolves over time. This makes Auto Mode agents easier to refine and standardize, especially for teams that maintain a shared library of skills across many agents.

How to access: Open an agent in Auto Mode Agent Builder, select an embedded skill to view or edit its instructions, then save or download it. No admin action is needed.

Learn more

Agent Builder [ROAD-1626]

Fathom connector available in public beta

You can now connect Fathom, the AI meeting assistant, to Glean. The new connector indexes meeting recordings, transcripts, summaries, action items, and related metadata, making the outcomes of your meetings searchable alongside the rest of your company knowledge. Glean respects each user's authorized Fathom access, so people only see the meeting content they are already permitted to view in Fathom. This helps teams find decisions, follow-ups, and discussion details from past meetings without switching tools or scrubbing through recordings.

How to access: In the Admin Console, go to Add data source and select Fathom, shown with a Beta badge, to configure the connector. Admin configuration is required.

Connectors [ROAD-1664]

Manager Insights: See Glean adoption across your team

Managers can now understand how their own organization is adopting Glean. Manager Insights gives eligible managers a scoped view of Glean adoption and usage across their reporting organization and direct reports, helping them spot where Glean is delivering value and where their teams may need enablement. The view is limited to each manager's own reporting line, while company-wide views and administrative controls remain available only to authorized Insights admins, so existing governance over usage data is preserved.

How to access: Eligible managers can open Insights in Glean to see adoption and usage for their reporting organization. No admin action is needed; company-wide views remain limited to authorized Insights admins.

Insights [ROAD-1609]

Agents tab in the browser extension

The Glean browser extension now includes an Agents tab, so you can run your agents from the extension side panel while browsing, without opening the Glean app. The Agents tab and its default-tab controls appear when browser-extension configuration is enabled for your deployment, and supported Glean Assist workflows include guidance to help teams migrate. This brings agents to the place many people already use Glean throughout the day, making it easier to trigger the right agent in the context of your current work.

How to access: Open the Glean browser extension side panel and select the Agents tab. Availability depends on browser-extension configuration, which is enabled automatically for eligible deployments; admins can manage rollout in extension settings.

Learn more

Browser Extension [ROAD-1461]

AI & Model Updates

Gemini 3.1 Pro Custom Tools available for Vertex AI

Admins using Vertex AI can now select Gemini 3.1 Pro Custom Tools as the large assistant model, paired with Gemini 3.7 Flash or Gemini 3.5 Flash as the small model. These options replace Gemini 2.5 Pro and Flash ahead of the October 6, 2026 routing change, giving Vertex AI deployments a supported upgrade path to the newer Gemini generation. The existing Gemini 2.5 and Claude pairings remain available until then, so you can plan the switch on your own schedule before the deadline.

How to access: Admins can select the new model pairing in Glean's model configuration settings.

Learn more

AI Models

Open model availability now managed by creator region

For deployments using Glean-managed keys, the availability of open models is now controlled by the creator's region rather than configured for each individual model. This simplifies model governance: instead of tracking many per-model choices, availability follows a single regional setting. Existing per-model choices migrate automatically, so nothing changes for your users and no admin action is required.

How to access: No action is needed; existing model settings migrate automatically for deployments using Glean-managed keys.

Learn more

AI Models

Agents move to the GPT-5.6 Luna model set

Agents on Glean-key migration plans now automatically move to the GPT-5.6 Luna model set where needed. This applies to published agents as well as unpublished drafts and staged edits, which receive the same migration without creating a new agent version. Explicit model selections revert to Glean's best-available choice, while all other agent behavior and configuration remain unchanged. This keeps agents on current, supported models without requiring builders to update each agent by hand.

How to access: No action is needed; the migration is applied automatically for agents on Glean-key migration plans.

Agents

Critical Updates

Review Gemini 2.5 model selections before October 6, 2026

Google is retiring Gemini 2.5 Pro, Gemini 2.5 Flash, and Gemini 2.5 Flash Lite on October 20, 2026. Glean stops routing to these models on October 6, 2026; after that, traffic moves to Gemini 3.1 Pro, Gemini 3.7 Flash with Gemini 3.5 Flash fallback where needed, or Gemini 3.5 Flash Lite. If your deployment currently uses any Gemini 2.5 model, review and update your model selections before October 6 to choose your preferred replacement rather than relying on automatic rerouting.

How to access: Admins should review model selections in Glean's model configuration settings before October 6, 2026.

Learn more

AI Models

Deprecations & Sunsets

Export to SharePoint artifact action retired

The Export to SharePoint artifact action has been retired because its underlying upload-from-URL capability is no longer available. You can still move artifacts into SharePoint: download the artifact from Glean, then upload it through the supported SharePoint workflow. No other artifact actions are affected.

How to access: Download artifacts from Glean and upload them to SharePoint using SharePoint's standard upload workflow.

Artifacts

Web SDK no longer supports the renderChat defaultSource option

The Glean Web SDK no longer supports the defaultSource option on renderChat. Embedded chat continues to work and uses your deployment-configured source defaults instead. If your integration passes this option, remove it from your SDK configuration; no other embedded chat behavior changes.

How to access: SDK integrators should remove the defaultSource option from renderChat calls; embedded chat otherwise continues to work unchanged.

Web SDK

Enhancements

Retry content-trigger activation after transient failures

When a temporary service issue prevents a content trigger from registering, activation now shows an actionable error and clears the failed registration behind the scenes. This means you can simply retry activation and succeed once the issue passes, instead of getting stuck with a trigger that will not activate.

How to access: No action is needed; if activation fails, retry it from the trigger setup flow.

Learn more

Agents

More reliable chat rendering while streaming

Chat messages now render reliably while responses are streaming. Previously, messages could briefly disappear or the layout could shift as content arrived. Conversations now stay stable and readable from the first token to the last.

How to access: No action is needed; the improvement applies automatically in Glean Chat.

Chat

Clearer audio cues in real-time voice sessions

Real-time voice sessions now use dedicated sounds for connecting, disconnecting, and tool activity, making it easier to understand what the assistant is doing at each moment. The tool-activity cue is quieter and no longer overlaps the assistant's voice, so responses stay easy to hear.

How to access: No action is needed; the new sounds apply automatically in real-time voice sessions.

Learn more

Assistant

Fewer blocked OAuth and MCP authentication flows

OAuth and MCP authentication and token-refresh requests now use a dedicated User-Agent. This reduces the chance that bot-based traffic filtering on external services blocks these flows, making connected tools and integrations more reliable to authenticate.

How to access: No action is needed; the change applies automatically.

Platform

Search and paginate static OAuth clients

Admins can now search and page through static OAuth clients directly from the OAuth settings page, making it much easier to find a specific client in deployments with many of them. Existing create, manage, and view-user actions are unchanged.

How to access: Open the OAuth settings page in the Admin Console and use the new search and pagination controls.

Admin Console

Dropdowns stay visible on smaller screens

Composer and model-picker dropdowns now shrink or flip direction to remain fully visible on screen. This improves usability on smaller windows and screens, where menus could previously extend out of view.

How to access: No action is needed; the improvement applies automatically.

Chat

MCP-created artifacts now appear only in Library

Artifacts created through Glean MCP, including from tools like Cursor, now appear only in your Library instead of also creating a Chat thread. This keeps your chat history focused on real conversations while artifacts created inside Glean Chat behave exactly as before.

How to access: No action is needed; find MCP-created artifacts in your Library.

Glean MCP

Authored request examples in Platform API documentation

Platform API documentation now provides authored, schema-valid request examples for the Agents, Chat, Skills, and Search APIs instead of generated placeholder payloads. Developers can copy realistic examples that work, shortening the path from reading the docs to a successful API call.

How to access: No action is needed; the improved examples appear in the Platform API documentation.

Developer Platform

Haptic feedback in the Glean mobile app

The Glean mobile app now provides subtle haptic feedback across key interactions: voice and dictation, chat and search, results, save, bookmark, pin and delete actions, filters, refresh, and failures. These small physical cues confirm your actions and make the app feel more responsive.

How to access: No action is needed; haptic feedback applies automatically in the Glean mobile app.

Learn more

Mobile App

Email trigger examples now match only your own mail

Gmail and Outlook trigger examples for labels, categories, and external senders now restrict matches to mail received by the trigger creator. This prevents triggers built from these examples from firing on coworkers' mail, keeping automations scoped to your own inbox.

How to access: No action is needed; the updated examples apply when you build new triggers from them.

Agents

Clearer tool deletion dialogs

Tool and tool-pack deletion dialogs now hide identifiers for agents you cannot access, keep links for agents you can see, and show an exact count of affected agents with guidance to contact an admin when needed. Deletion terminology is now consistent throughout.

How to access: No action is needed; the improved dialogs appear when deleting tools or tool packs.

Admin Console

Outdated onboarding promos removed from the extension

Outdated chat and search onboarding promos have been removed from the browser extension side panel, reducing clutter and keeping the panel focused on your work.

How to access: No action is needed; the change applies automatically in the browser extension.

Browser Extension

Refreshed indexed-items chart in data source setup

The indexed-items bar chart in data source setup now uses current design-system colors while preserving its existing labels and hover text, making the setup experience more consistent with the rest of the Admin Console.

How to access: No action is needed; the updated chart appears in data source setup.

Admin Console

Content-trigger filters recognize user fields

Content-trigger facet pickers now recognize user fields from data source schemas, including a current-user option and people icons. This makes it easier to build triggers that react to documents based on the people associated with them.

How to access: No action is needed; the options appear automatically when configuring content-trigger filters.

Learn more

Agents

Search skills by name when adding to an agent

The Add existing skill dialog in Agent Builder now lets you search personal, shared, and platform skills by name, so you can quickly find the right skill instead of scrolling through the full list.

How to access: In Agent Builder, open the Add existing skill dialog and type a skill name to search.

Agent Builder

Corrected French label for Chat settings

The French label for Chat settings now displays as Paramètres du chat instead of falling back to English, improving the experience for French-language users.

How to access: No action is needed; the corrected label appears automatically for users with French language settings.

Chat

Full web crawls cover every configured sitemap and seed URL

Web crawler full crawls now enqueue every configured sitemap and seed URL, ensuring complete coverage of the web content you have configured. Incremental crawl behavior is unchanged.

How to access: No action is needed; the improvement applies automatically on the next full crawl.

Connectors

Clearer guidance for Slack send-message actions

Slack send-message actions now validate channel IDs and provide clearer guidance when a channel is archived, inaccessible, externally shared, or an unavailable private channel. This helps agents and users understand why a message could not be sent and what to do about it.

How to access: No action is needed; the improved validation applies automatically to Slack send-message actions.

Actions

Restrict LLM models by department and IdP group together

Admins can now restrict access to LLM models by department and IdP group at the same time, enabling more precise model governance. For example, you can make a specific model available only to certain departments within a particular identity-provider group.

How to access: Configure the combined restrictions in the model access settings of the Admin Console.

Learn more

Admin Console

Bring your own Deepgram API key for Meeting Notes

Customer administrators can now configure a customer-owned Deepgram API key for Meeting Notes transcription on supported deployments. This gives organizations direct control over their transcription provider relationship, billing, and usage.

How to access: In the Admin Console, go to Assistant → Settings → Meeting Notes and enter your Deepgram API key. Available on supported deployments.

Admin Console

Failed connector configurations retained for 14 days

Failed connector configuration attempts are now retained for 14 days instead of 3. This gives admins more time to troubleshoot and retry a failed setup without re-entering credentials and secrets. No customer action is required.

How to access: No action is needed; the longer retention applies automatically.

Connectors

Strengthened workload isolation with non-root containers

Many Kubernetes workloads that power Glean now run as non-root users, strengthening workload isolation and reducing the potential impact of any single compromised component. This security hardening happens entirely behind the scenes.

How to access: No action is needed; the hardening applies automatically.

Platform

Safer Azure Files permission enforcement

Azure Files permission enforcement now reconciles shared New Technology File System (NTFS) permissions safely, handles custom access-control entries, and rejects malformed security descriptors. This keeps search results for Azure Files content accurately aligned with your file-system permissions.

How to access: No action is needed; the improvements apply automatically to the Azure Files connector.

Connectors

Action usage metadata respects agent and workflow permissions

Action usage metadata now follows agent and workflow permissions: details you can access remain visible, metadata for inaccessible items stays hidden, and unlisted workflows are omitted. Deletion behavior is unchanged. This ensures usage views never reveal information about agents or workflows you are not permitted to see.

How to access: No action is needed; the permission enforcement applies automatically.

Actions

Consistent permission enforcement for MCP servers using Glean Skill tools

Customer-created MCP servers that use Glean Skill tools now consistently enforce platform audience permissions across skill discovery, skill-file disclosure, and tool execution. Restricted tools can no longer be exposed or invoked through lazy disclosure, keeping your skill permissions airtight when accessed through MCP.

How to access: No action is needed; the enforcement applies automatically to customer-created MCP servers.

Glean MCP

Updated permissions for Dataproc Serverless 3.0

The Dataproc runtime custom role now includes the permissions required for Spark jobs on Dataproc Serverless 3.0. Managed deployments require no admin action.

How to access: No action is needed for managed deployments; the updated role applies automatically.

Platform

Better retry behavior under rate limiting

Rate-limit responses now preserve Retry-After information and propagate it when a request is ultimately denied. API clients and integrations can use this to back off intelligently and retry at the right time instead of guessing.

How to access: No action is needed; clients should honor the Retry-After header when retrying throttled requests.

Developer Platform

More reliable Ironclad syncing under rate limits

Ironclad admin and user-token listings, attachment downloads, and permission dispatch now use endpoint-scoped queues aligned with Ironclad's organization rate limits. This makes Ironclad crawling and permission syncing more reliable for deployments with heavy Ironclad usage.

How to access: No action is needed; the improvement applies automatically to the Ironclad connector.

Connectors

Faster, more reliable document processing scheduling

Parsing-service replicas now fit standard node sizes, improving scheduling and avoiding pending pods during document processing. This keeps indexing pipelines moving smoothly during heavy processing periods.

How to access: No action is needed; the improvement applies automatically.

Platform

Cleaner AI-security log exports

AI-security log exports no longer include the internal GUARDRAIL_CONTEXT payload, producing cleaner exports for downstream analysis. Guardrail detection and enforcement behavior are unchanged.

How to access: No action is needed; the change applies automatically to AI-security log exports.

Protect

Fixes

Content-trigger people filters return resolvable users

Content-trigger people filters now return resolvable user values, so the users you select produce firing conditions reliably instead of silently failing to match.

Agents

Memory and personalization available for more users

Memory and personalization are now available for users whose account details previously could not be resolved, using an additional account source to identify them correctly.

Assistant

Retired ServiceNow news articles removed from search

Retired ServiceNow news articles no longer appear in Glean search results because inactive records are filtered out.

Connectors

Deactivated ServiceNow knowledge articles removed from search

Deactivated ServiceNow knowledge articles no longer appear in Glean search results.

Connectors

Prompt cleanup of deactivated ServiceNow articles

Deactivated ServiceNow knowledge articles are now removed promptly during incremental crawls instead of waiting for a full crawl, so stale content disappears from search much sooner.

Connectors

Original source

This release expands Glean’s AI capabilities, improves agent administration and collaboration, adds new platform integrations, and strengthens security, privacy, and reliability across connected data sources.

Major Launches

Connect Glean content events to your automations

The experimental Glean Triggers API lets developers subscribe to content events from connected data sources and deliver those events to external platforms and agents. You can register trigger subscriptions, search matching events, manage a trigger’s lifecycle, and receive signed webhook deliveries. This makes it possible to connect Glean activity to automation platforms, developer tools, and custom agent workflows while preserving the permissions of the subscribing user. The API is designed for integration builders who want Glean events to initiate work outside the Glean interface.

How to access: Use the Platform API with a token that has the required trigger scope, and include the experimental opt-in header on requests. The API is intended for developers and integration builders; the surface is experimental and may change.

Developer Platform

See your AI security posture in one place

Guardrails Insights gives security and governance teams a centralized view of AI guardrail activity. Instead of reviewing individual findings one at a time, administrators can use the dashboard to understand trends, policy confidence, topics, and high-risk users. The result is a clearer picture of where AI usage is creating risk and where investigation or policy changes may be needed. This is especially useful for teams responsible for monitoring AI usage, reviewing violations, and communicating security posture across the organization.

How to access: Open the AI Security Guardrails Insights dashboard from Glean Protect. Access depends on your organization’s Protect configuration and administrator permissions.

Protect

Preserve important agent drafts with saved versions

Agent Builder now lets you create named, immutable checkpoints of a draft agent. Saved versions give you a dependable way to preserve a meaningful configuration before making further edits, without changing the existing draft, staged, and published lifecycle. You remain in the editable draft state after saving. When you want to reuse an earlier checkpoint, select it from version history and load it back into draft, then continue editing or publish it through the normal workflow. This helps teams experiment more safely and revisit prior working states without maintaining parallel copies.

How to access: Open an agent in Agent Builder and use the version history or version selector to save, view, load, or delete draft checkpoints.

Agents

Rediscover useful work in the redesigned Library

The redesigned Library makes durable artifacts easier to find and revisit. You can browse artifacts you created, artifacts shared directly with you, and explicitly public company artifacts. New filters help you narrow results by creator team or department, while sorting options make it easier to find recently created or popular content. The experience is designed to move beyond a constantly changing recently viewed list, so high-signal artifacts are easier to rediscover after the original task is complete.

How to access: Open Library in Glean and use the ownership, department, creator, and sorting controls to browse and filter artifacts.

Library

Bring support agents into your browser workflow

Agents in the Glean Browser Extension lets users launch configured support agents directly from supported Zendesk, Salesforce Service Cloud, and ServiceNow pages. The agent can use the current ticket or case context so you can work without copying information into another window. This provides a replacement path for embedded support experiences and gives administrators a dedicated way to configure which agents are available. Starting August 24, 2026, the experience is generally available to eligible deployments with existing embedded-integration usage.

How to access: Install and sign in to the Glean Browser Extension. Eligible users can open the extension sidebar on a supported ticket or case page and select the Agents tab. Admins can configure access from the browser-extension settings.

Learn more

Agents

AI & Model Updates

Choose Gemini 3.7 Flash for fast, capable agent work

Gemini 3.7 Flash is now available as a model option in Assistant and Agents. It is designed for high-volume, low-latency work while supporting stronger reasoning for well-scoped tasks. You can select low, medium, or high thinking levels depending on whether you want faster responses or more deliberate reasoning. The model is available across supported Assistant and Agents surfaces, with access varying by deployment key configuration. Gemini 3.6 Flash remains available under More models, so existing model choices are not removed.

How to access: In Assistant, open the model selector and choose Gemini 3.7 Flash. For Customer Key deployments, an administrator may need to make the model available from Admin Console → Platform → Models.

Assistant and Agents

Get smarter defaults for workflow agents

Workflow agents that use “Let Glean decide” now default to GPT 5.6 Luna when it is available, with GPT 5.1 used as a fallback. This improves the default model choice without requiring builders to update existing agents. Explicit model selections remain unchanged, so teams that have deliberately chosen a specific model keep that configuration. The change applies to the default path only and is intended to improve the baseline experience for workflows that do not specify a model themselves.

How to access: No action is required for existing workflow agents using “Let Glean decide.” Open an agent’s model settings in Agent Builder if you want to review or choose an explicit model.

Agents

Use Anthropic models with your custom key

Custom-key customers can now use supported Anthropic models through Glean for production workloads. This gives teams another provider option when selecting models for their organization’s AI experience and lets customers align model usage with their existing provider and key-management strategy. Availability depends on the models supported by the configured Anthropic provider and the deployment’s permissions and setup.

How to access: An administrator must configure the Anthropic provider and custom key in the model settings available to your deployment. Once enabled, users can select supported Anthropic models wherever the model picker is available.

Models

Use real-time voice more naturally when you are interrupted

Real-time voice now includes a Talk uninterrupted control when Glean detects that you started speaking while it was responding. Tap the notice to switch into a push-to-talk interaction and finish your turn before Glean responds. This gives you a clearer way to continue speaking without losing your thought or competing with the assistant’s response. The control is designed for natural back-and-forth conversations on web, desktop, and mobile, where interruptions are expected and a hands-free interaction should not require restarting the exchange.

How to access: Start a real-time voice session from the waveform button in the composer. If Glean detects an interruption, tap the Talk uninterrupted notice to continue speaking.

Learn more

Assistant

Critical Updates

Review the new Engram Memory billing treatment

Engram Memory usage is now billed for Glean Core Suite plans across supported models. Migration usage remains free. For Enterprise Flex credit plans, usage remains free through September 15, 2026, and begins consuming FlexCredits on September 16, 2026. This change affects how memory-related usage appears in plan economics, but it does not change the underlying memory experience. Review your plan and usage expectations before the Enterprise Flex billing date if your organization relies on persistent Assistant memory.

How to access: No product action is required. Administrators should review their plan and usage reporting, and contact their Glean representative with billing questions.

Assistant Memory

Update integrations that use the Agents subscription schema

The public Agents REST subscription schema no longer includes the allowAllBgWriteActions field. Integrations that referenced this field should remove it from their requests or generated models. Background-run behavior and the internal setting remain unchanged, so this is a schema cleanup rather than a change to how subscriptions execute. The update matters only to integrations that directly consume the public subscription schema and expect the removed field to be present.

How to access: Update any integration or generated client that references allowAllBgWriteActions, then validate the subscription request against the current public schema.

Assistant API

Require HTTPS for dynamically registered MCP servers

Dynamic MCP client registration now requires server URLs that use HTTPS. Registrations that use an http:// URL will fail with a clear error instead of sending registration traffic over plaintext. This protects credentials and registration metadata while preserving the existing dynamic-registration workflow for secure servers. If you manage a custom MCP server that uses dynamic registration, confirm that its public registration URL is HTTPS before attempting to connect it.

How to access: No action is needed for existing secure servers. Update any custom MCP server configuration that uses an http:// registration URL to use HTTPS.

MCP

Limit certificate setup to full administrators

Certificate management APIs now require full administrator permissions to complete the SSL certificate setup flow. This narrows a sensitive configuration path to the role intended to manage organization-wide certificates and reduces the chance that a lower-privilege administrator changes security infrastructure unintentionally.

How to access: Full administrators can continue to complete certificate setup from the administrator experience. Users without full administrator permissions should contact an administrator with the required role.

Admin Console

Keep OneDrive access groups consistent

OneDrive Product Access Group membership checks now use stable datasource user IDs. This prevents inconsistent access decisions caused by relying on identifiers that could vary between checks. Administrators should see more reliable enforcement of the groups that control which users can access OneDrive content through Glean, without needing to change the groups themselves or reconfigure the connector.

How to access: No action is required. Continue managing Product Access Groups through the existing OneDrive administration workflow.

OneDrive

Let users request access to protected chats and artifacts

Users who open a protected chat or artifact they cannot access can now submit an access request instead of relying on an out-of-band message. The requester can optionally include context, while the owner can review and approve or deny the request. Approved users join under the existing participant and permission controls. This creates a clearer, auditable path for collaboration while keeping the owner or authorized administrator in control of access.

How to access: Open a protected chat or artifact. If you do not have access, use the request-access action and optionally add a message for the owner.

Assistant

Give Viewer-shared chats a consistent home in Projects

Chats shared with Viewer access now appear in Projects, including deployments where Multiplayer Chat is disabled. This makes shared conversations easier to find and gives recipients a consistent place to return to chats they can read. The change does not expand the underlying permission: a Viewer can see the chat only when it has been shared with them and continues to have Viewer-level access.

How to access: Open Projects and look for chats that have been shared with you as a Viewer.

Projects

Deprecations & Sunsets

Use the updated Embedded integrations administration URL

The Embedded integrations administration page now uses a URL that matches its page name. Existing Glean Assist links and bookmarks continue to work through redirects, so saved links do not need to be changed immediately. When updating internal documentation, onboarding materials, or bookmarks, use the new Embedded integrations route so future references align with the current product terminology and the browser-extension transition.

How to access: Navigate to Embedded integrations from the Admin Console. Existing Glean Assist URLs will redirect to the updated page.

Embedded Integrations

Enhancements

Find skills faster in Settings

The Skills tab in Settings now includes search by skill name. Use it to filter personal, shared, and platform skills instead of scanning the full list. Search results update within the current tab and show a clear empty state when no skill matches. This is particularly useful for organizations with many shared skills or for users who remember the purpose of a skill but not where it appears in the list.

How to access: Open Settings → Skills and enter a skill name in the search field.

Skills

Understand agent tools at a glance

Agent Builder now separates datasource tools into Read and Write sections. Each section has its own count, search behavior, and Add all tools control, making it easier to understand what an agent can retrieve and what it can change. Approval settings appear only when write tools are selected, so builders can see the relevant safety control in context instead of interpreting a mixed list. The same distinction is available when configuring action packs and MCP servers.

How to access: Open an agent in Agent Builder and expand a datasource, action pack, or MCP tool selection. Use the Read and Write sections to review or change the toolset.

Agent Builder

Make agent version history easier to scan

The version history menu now puts saved-version names first, with the date and author underneath. Only the current version is labeled Published, and empty histories explain how to begin tracking changes. Expanded version groups scroll into view, and hover highlights clear correctly after selection. Together, these changes make it easier to distinguish versions, understand the current lifecycle state, and navigate a growing history without opening each version one by one.

How to access: Open an agent in Agent Builder and open its version history menu.

Agents

Start voice with useful context sooner

Voice onboarding is now shorter and more personalized. New users see a concise welcome and can opt into a daily briefing rather than stepping through a multi-screen tutorial. The briefing can help users begin with relevant work context, while returning users continue to receive the standard voice greeting. This reduces setup friction and gets you to a useful voice interaction faster without changing the normal experience after onboarding.

How to access: Start real-time voice from the waveform button in the Glean composer. New users will see the updated welcome automatically.

Learn more

Assistant

Identify inactive agents before cleanup

The inactive agents CSV export now includes a Last edited by email column. Administrators can use this information to identify the person who last changed an unused agent before deciding whether to archive, update, or remove it. The added detail makes cleanup conversations more targeted and reduces the need to investigate ownership through separate systems.

How to access: Export the inactive agents CSV from the Agents administration or insights experience and review the new column.

Agent Insights

Preserve approval settings while editing tool selections

Agent Builder now retains Run without approval settings when you edit actions in a selected action pack or change the tools selected for an MCP server. This prevents an edit to the tool list from unexpectedly resetting an intentional approval choice. Newly added action packs and MCP servers still start with Run without approval turned off, so builders must explicitly enable the setting for new additions.

How to access: Open an agent’s action pack or MCP configuration in Agent Builder, then edit its tools. Review the approval setting before saving changes.

Agent Builder

Complete support-agent forms with less clutter

Support agents now show automatically captured ticket-page content as a compact, non-editable pill instead of placing the full captured content directly in the form. The complete context is still sent when the agent runs. Forms also place focus on the first empty field, so you can begin entering the information the agent actually needs without moving past prefilled context. This makes browser-based support workflows easier to scan and faster to complete.

How to access: Open a supported Zendesk, Salesforce Service Cloud, or ServiceNow page with the Glean Browser Extension, then open the Agents tab.

Learn more

Glean Extension

Make agent status notifications clearer

Agent Builder notifications now state the outcome of the action that occurred. Publishing, sending an agent for review, and approving or rejecting a review each display a corresponding success message. Import failures are now reported as errors instead of warnings, making it easier to distinguish a completed action from a problem that needs attention. These clearer messages reduce uncertainty during publishing and review workflows, especially when several changes are being made in sequence.

How to access: No action is required. The updated messages appear automatically in Agent Builder when you publish, review, approve, reject, or import an agent.

Agent Builder

Search large app lists more easily

The Agent Builder app-selection dropdown now supports search. When many Teams or Slack apps are available, enter an app name to narrow the list instead of scrolling through every option. This reduces the time required to configure an agent for a specific app and makes the selection experience more manageable for organizations with multiple connected workspaces or channels.

How to access: Open the app-selection dropdown in Agent Builder and enter the name of the Teams or Slack app you want to select.

Agent Builder

Run eligible MCP tools in the background

MCP tools now support background execution by default. This allows eligible tool calls to continue without blocking an interactive flow, while administrators retain control through each tool’s Background Execution Policy. You can disable background execution when a tool needs to remain synchronous or requires closer review. The setting is applied per tool, so teams can choose a different execution policy for different connected capabilities.

How to access: Administrators can review each MCP tool’s Background Execution Policy from the MCP server configuration and disable background execution when needed.

MCP

Allow eligible Composio writes to run without a confirmation click

Administrators can now allow Composio write tools to run in the background without manual confirmation by default. Existing policies remain unchanged, and administrators continue to control the behavior through each tool’s Background Execution Policy. This can shorten workflows where a write is already governed and does not need an additional confirmation step, while preserving the ability to require confirmation for higher-risk actions.

How to access: Configure the Background Execution Policy for the relevant Composio write tool. Agent creators can then use the eligible action in an interactive agent.

Actions

Separate host-app setup from MCP server management

The Glean MCP servers administration page now opens with separate Host apps and MCP servers tabs. Host apps show where connected clients are configured and used, while MCP servers provide the server and tool-management workflow. Separating these concepts makes it easier for administrators to find setup information, review usage, and manage the right object without navigating through a combined page.

How to access: Open the Glean MCP servers page in the Admin Console and choose the Host apps or MCP servers tab.

MCP

Add Skills tools to selected MCP servers

Administrators can add Glean’s Skill tools—find_skills, read_skill_files, and run_tool—to an MCP server. Connected MCP clients can then discover relevant skills on demand and run their underlying tools without loading the full catalog up front. The tools are opt-in per server and are not enabled automatically, giving administrators control over which clients can discover and execute skills through each MCP connection.

How to access: Open an MCP server in the Admin Console, choose the Skill tools pack from the tool catalog, and save the server configuration.

MCP

Find Glean content through DocuSign MCP

DocuSign MCP is now generally available through Glean’s managed MCP catalog using DocuSign’s production server and OAuth endpoints. This gives administrators a production connection path for DocuSign workflows through the managed catalog. Connections created during the beta period may still point to the earlier environment, so they may need to be recreated or re-authenticated before they use the production endpoints.

How to access: Open the managed MCP catalog, select DocuSign, and complete the OAuth connection flow. If you connected during beta and experience an issue, recreate or re-authenticate the connection.

Docusign MCP

Make streamed chat responses feel smoother

Chat responses now use a subtle fade-in animation as new text streams into the conversation. The change keeps the existing generation indicator while making the response feel less abrupt as it appears. It does not change the answer content, model behavior, or response timing; it is a presentation improvement intended to make active generation easier to follow and the overall chat experience more polished.

How to access: No action is required. The animation appears automatically while new Assistant text streams into a chat.

Assistant

Give OAuth clients authorized access to Skills APIs

Dynamically registered OAuth clients can now request the Skills scope when they need authorized access to Glean’s Skills APIs. This provides a more precise permission boundary for integrations that discover or use skills, without requiring broader access than the client needs. Existing clients are unaffected unless they need Skills API access; those clients should update their registration request and consent configuration to include the new scope.

How to access: Update the dynamic OAuth client registration to request the Skills scope, then complete the normal authorization flow.

Skills API

Resolve skill access more efficiently

Skill access checks now use an optimized permission filter by default. The change improves how Glean resolves which skills are available to a user while preserving the intended access rules. Users and administrators should see the same permission outcomes, with a more efficient path used behind the scenes when Glean evaluates accessible skills.

How to access: No action is required. Skill availability continues to follow your existing permissions and sharing settings.

Skills

Match exported LLM token counts to billing

Exported LLM event data now reports cache-adjusted input-token counts, matching the values used for billing and Agent event exports. This prevents the raw and cache-related token buckets from overlapping and making manual cost calculations appear higher than the billed amount. The change does not alter the underlying cost calculation; it makes the exported data reflect the same adjusted input count used by pricing.

How to access: No action is required. Use the updated input-token count in exported LLM event data when performing manual usage or cost analysis.

LLM Usage

Fixes

Reduce false positives when reviewing pasted or quoted text

AI Security guardrails now better distinguish a genuine prompt-injection attempt from text you are pasting, quoting, or reviewing. This reduces false positives for configuration files, security documentation, assistant responses, and ordinary business questions while preserving detection for concealed or agent-directed attacks. Findings also provide more useful investigation context by showing the triggering text, its source and location, and the potential impact on an agent.

Protect

Keep ServiceNow search text out of request URLs

ServiceNow search inputs are now sent in the request body rather than the URL. Search behavior and supported inputs remain unchanged, while search text and related parameters are less likely to be retained in load-balancer access logs. This reduces exposure of potentially sensitive business information and personal data during normal ServiceNow searches.

ServiceNow

Prevent GitLab access changes from removing valid memberships

GitLab identity updates now preserve crawl-scope information so memberships populated by a dedicated crawl are not incorrectly removed by a simple identity run. This addresses a race condition that could cause some users to lose access to GitLab documents and makes access updates more reliable when multiple identity crawls are active.

GitLab

Index Notion pages that previously disappeared from search

Some Notion pages could be silently omitted from the search index when an internal document-processing path encountered missing chunk metadata. The parser now handles that case safely, allowing affected pages to be indexed instead of being dropped. This restores discoverability for pages that were present in Notion but did not appear in Glean search.

Notion

Keep SharePoint crawling reliable with many custom columns

SharePoint site-page crawling now selects only the fields required by Glean when custom columns are not configured. This avoids requests that fail when a list contains a large number of unsupported custom columns, allowing the crawl to continue and making site-page content more reliable to index.

Sharepoint

Original source

AI & Model Updates

Get faster, more capable responses with Gemini 3.7 Flash

Gemini 3.7 Flash is now available as a model option in Glean Assistant and Agents. It is designed for high-volume, tool-using workflows where response speed matters, while providing stronger performance for well-scoped reasoning and coding tasks. The model is available across Glean Universal Model Key and Customer Key deployments covered by the supported release version. Gemini 3.7 Flash appears as a premium model, while earlier Flash models remain available in the model picker for teams that need them.

How to access: Open the model selector in Assistant or choose a model when configuring an agent. Administrators can exclude the model from Admin Console → Platform → Models if their organization does not want it available.

Assistant & Agents [ROAD-1647]

Improve Fast Mode responses with GPT-5.6 Luna

Fast Mode now uses GPT-5.6 Luna with low reasoning instead of the previous lightweight model. You should see improved answer quality while retaining the quick response experience that Fast Mode is designed to provide. No workflow changes are required: existing Fast Mode conversations and entry points continue to work, and the model change is applied automatically as the updated default becomes available.

How to access: Select Fast Mode in Assistant as usual. No administrator or end-user action is required.

Assistant

Give admins control over agent default models

Administrators can configure the default model set used by agent steps, giving teams more control over the models selected across agent types and supported modes. This helps organizations establish a consistent baseline for agent behavior instead of relying only on product defaults. The setting applies where the selected mode supports the configured model set; individual agent or step configuration can continue to determine behavior where those controls are available.

How to access: Administrators can use the deployment’s agent model configuration controls to set the default model set. No action is required from end users unless their administrator changes the organization’s defaults.

Agents

Enhancements

Open chat side panels without text reflow

Chat side panels, including artifact and source panels, now open and close with smoother transitions. The panel reaches its final width immediately while its contents animate in place, so the conversation no longer repeatedly resizes or rewraps as the panel appears. Closing a panel also returns you directly to the main conversation view without intermediate layout shifts. This makes it easier to keep your place while reviewing sources, working with artifacts, or moving between the conversation and supporting content on web and mobile.

How to access: Open an artifact or chat source panel from a conversation. The improved behavior is automatic; no settings change is required.

Chat

Control staged agent review request emails

Moderators can now opt out of staged agent review request emails through their collaboration email notification settings. This gives moderation teams more control over review-request traffic while preserving the messages needed to keep agents reviewed.

How to access: Update collaboration email preferences for moderation notifications in your notification settings. No action is needed if you want to keep receiving these messages.

Notifications

See agent risk posture more clearly in Agent Scan

The Agent Scan experience now gives administrators a clearer view of agent risk. The dashboard separates Critical and High risk counts, combines Medium and Low risk agents into a simpler summary, and exposes the counts needed for the risk overview. Filters are also easier to use: Creator is a compact searchable single-select control, Risk supports multiple selections, and policy status styling is consistent across the admin experience. Policy audiences now show names for individual people or groups and reveal the full recipient list on hover.

How to access: Open the admin Agent Guardrails page and use the Agent Scan dashboard, filters, and policy audience details. No end-user action is required.

Protect

Understand agent permissions before activation

Before you activate an agent, the permissions screen now explains what each tool can do and whether it may act without asking for approval. Tools are identified as read-only, able to run without approval, or requiring approval. MCP servers also show whether all, some, or none of their enabled write tools can run unattended. The display follows administrator approval policies, so a tool that an administrator requires to pause will be shown as requiring approval even when the agent’s own configuration requests otherwise.

How to access: Open an agent that uses tools or MCP servers and review the permissions screen before activating it. No configuration change is required.

Agents

Separate read and write tools for MCP servers

When you build or review an agent, each MCP server’s capabilities are now separated into Read tools and Write tools with individual counts. This makes it easier to see whether a server can only retrieve information or can also change data. The layout also makes the write-tool approval control easier to understand, and the control is hidden when a server has no enabled write tools. Selecting or removing a tool in one section does not change the selections in the other section.

How to access: Open an agent’s Capabilities tab and expand an MCP server. Review or adjust the Read tools and Write tools selections there.

Agents

Share meeting notes and imported summaries directly from the artifact

Artifact owners can now use the Share action directly from the header of meeting notes and imported meeting summaries, even when the original chat thread is no longer available. This gives owners a reliable way to share the artifact itself without needing to reopen or locate the conversation that created it. In cases where no linked chat thread exists, the Chat action remains hidden so the available actions accurately reflect what you can use.

How to access: Open a meeting note or imported meeting summary that you own, then select Share from the artifact header.

Artifacts

Retrieve schemas for Glean-native tools through the Tool Metadata API

Clients can now retrieve input schemas and descriptions for Glean-native tools through the Tool Metadata API. This helps integrations understand the inputs expected by native capabilities such as Meeting Lookup before invoking them. Existing action-pack and external MCP behavior remains unchanged, so current integrations do not need to change how they use those tools. The improvement is especially useful for clients that need to inspect tool definitions dynamically while building or running agent workflows.

How to access: Use the Tool Metadata API when your client needs the input schema or description for a Glean-native tool. No end-user action is required.

Agents

Restore full insights visibility for Insights Moderators

Insights Moderators can again view insights for all agents, including agents they do not own or edit. This restores the broader visibility needed by moderation and oversight teams to review agent activity across an organization rather than being limited to agents associated with their own accounts. Agent owners and regular users do not need to change their sharing settings for this access to work.

How to access: Sign in with an Insights Moderator role and open the agent insights experience. Access is controlled by your organization’s role assignment.

Insights

Keep MCP moderation focused on Glean-hosted servers

MCP Server Moderators are now limited to administering Glean-hosted MCP servers and viewing MCP Insights. Managing third-party MCP connections—including creating, updating, deleting, refreshing, or repairing third-party hosts and accessing their templates—continues to require the existing tool-administration permissions. This keeps moderation responsibilities aligned with the surfaces they are intended to govern while preserving established controls for external connections.

How to access: MCP Server Moderators can use the Glean-hosted MCP administration and MCP Insights surfaces. Ask an administrator for tool-administration permissions when third-party MCP management is required.

MCP

Find recent Outlook mail sooner

New or changed Outlook messages delivered through webhooks are now prioritized ahead of background full-refresh crawls in the mailbox thread queue. As a result, recent mail can become searchable sooner without interrupting other mailbox crawl operations. This is an automatic indexing improvement: it does not change how users search Outlook, and it does not require administrators to reconfigure the connector.

How to access: Search Outlook content in Glean as usual. No user or administrator action is required.

Outlook

Improve OneDrive and SharePoint crawl reliability

OneDrive and SharePoint crawls now use more efficient exact-key batch deletion when cleaning up delta results. This reduces the chance of slowdowns or stalls on large drives and helps keep connector processing predictable as content changes over time. The improvement is applied to the crawl process automatically and does not change the way users search or access content.

How to access: Continue using OneDrive and SharePoint connectors normally. No user or administrator action is required.

OneDrive & SharePoint

Capture user intent for MCP usage analysis

MCP tool calls now capture the user’s intent for usage logging and downstream analysis. This gives organizations better context for understanding why MCP tools are used, supporting more useful operational and adoption analysis. The change does not alter the tool’s result or require users to provide a new form field outside the MCP interaction. Intent capture is handled as part of the MCP tool-call experience and its associated usage logging.

How to access: No action is required. Intent information is captured automatically for supported MCP tool calls.

MCP

Fixes

Hosted Excel, Outlook, SharePoint, and Teams actions are more reliable

Hosted Composio actions for Excel, Outlook, SharePoint, and Teams are now more reliable. An infrastructure issue could previously cause intermittent failures when these actions were used in hosted workflows. Reliability protections have been improved so the supporting service is less likely to become unavailable during periods of platform pressure. Existing workflows continue to work without changes, and no action is required from users or administrators.

Actions

ServiceNow Create Request forms load correctly

ServiceNow Create Request no longer fails with “Failed to fetch action data” after you select a catalog when the action has no catalog restriction configured. The form now proceeds to the catalog-item step when no specific allowlist is set, while configured catalog restrictions continue to be enforced. This restores the expected request-creation flow without requiring administrators to add an unnecessary catalog restriction.

ServiceNow

Sign in to the Android app with Intune MAM policies

Android sign-in now works more reliably for organizations using Intune Mobile Application Management, including BYOD devices that are not enrolled in full device management. The flow handles both permissive and restrictive “Receive data from other apps” policies, uses Microsoft Edge when Intune requires a managed browser, and provides clearer guidance when Edge is missing or not recognized as managed. Users without Intune app protection are unaffected.

Learn more

Glean Mobile

Find identity-provider groups when sharing

Searching for identity-provider groups, such as Azure AD or Microsoft 365 groups, is now more reliable when sharing an agent, chat, or other item. You can search using text from any part of a group name, and matching groups continue to appear consistently in tenants with many groups. Previously, a group could be omitted unless you typed the full name or started with its exact prefix. The sharing flow and group selection behavior remain unchanged once you find the group.

Sharing

Export and import agents without tool-creation permissions

Agent export and import now work for users who do not have permission to create tools when the agent uses datasource actions. Previously, the export could fail because an action could not be resolved to a portable name. Actions are now represented with stable names rather than environment-specific identifiers, allowing the workflow to be downloaded and reused correctly without granting additional tool-creation permissions.

Agents

Original source

New cadence

We’re moving to twice-weekly release notes to match our release cadence, so you’ll know what shipped sooner.

✨ Major Launches

Find and reuse your past Assistant chats without leaving the conversation

You can now ask Glean to find and pick up your earlier Assistant conversations directly from a new chat, so you can continue long-running work without hunting through your history. Only your own conversations that are stored in Chat History are surfaced, and existing retention and privacy controls still apply.

How to access: In a new Assistant conversation, just ask Glean to find a previous chat (for example, reference what you were working on). Glean surfaces matching conversations from your own Chat History. No setup is required.

Assistant

Discover useful next steps directly from Glean Home

Glean Home can now surface personalized Cowork suggestions based on your recent work. Each card recommends a concrete next step, such as preparing for a meeting, summarizing a call, drafting a follow-up, or updating a document, and opens the relevant Assistant workflow with context already provided. This helps you discover valuable Glean capabilities without knowing the exact prompt or navigating through multiple product areas. Suggestions are designed around practical work journeys and include source-backed context so you can understand why a recommendation is relevant.

How to access: Open Glean Home and select a suggested card when one appears. No admin setup is required for supported Glean Key deployments.

Glean Home [ROAD-1444]

Understand usage by department and support internal chargeback

Billing administrators can now see how Glean usage is distributed across departments. The new department view helps you compare consumption, identify teams that are under-adopting or driving the most usage, and export the data for budgeting, chargeback, showback, and internal reporting. Department attribution follows the department metadata available through your identity provider or organization chart, so the view reflects the structure your organization already maintains. This gives finance and platform teams a clearer organizational lens on Glean adoption and consumption.

How to access: Open Admin Console → Usage and select By departments. Use the department table to compare usage or download the data as a CSV.

Learn more

Usage and Billing [ROAD-1340]

Set monthly usage limits for individual departments

Billing administrators can now set a monthly usage limit for each department. This lets you align Glean usage with the way your organization manages budgets, give different teams different levels of headroom, and choose between alert-only monitoring and blocking additional billable usage after a limit is reached. Department limits apply to each member of the department rather than creating a shared pool, and individual user overrides remain available for people who need more capacity. Limits reset at the start of each calendar month.

How to access: Open Admin Console → Usage → By departments, open the department’s actions menu, select Manage usage limit, enter the monthly amount, and save. Confirm that department metadata is mapped in your people data.

Learn more

Usage and Billing

Delegate work to dynamic sub-agents with Task Tool

Autonomous Agents can now delegate work through the Task Tool by default in supported workflows. This gives an agent a consistent way to create and route work to dynamic sub-agents, support follow-up steps, and use model routing suited to the task. You can use the capability to break a larger workflow into focused pieces while preserving the agent’s permissions and error-handling behavior. The change also removes the need for the previous external experiment path for supported Task Tool workflows.

How to access: Create or open a supported autonomous agent and configure a step that delegates work through the Task Tool. The capability is enabled automatically for supported workflows.

Autonomous Agents [ROAD-1622]

Let agents act as themselves with governed service credentials

Agent Identity lets an agent use its own scoped service credentials instead of borrowing the identity of the person who invoked it. Actions can therefore appear under the agent’s own account in connected systems, while Glean records the triggering user or schedule separately for auditability. Admins can limit the credential to the tools and data the agent needs, reducing permission sprawl and preventing the agent from inheriting unintended user access. Scheduled and background agents can also continue running without depending on a person’s OAuth session.

How to access: An admin creates and scopes a service credential in the target system, registers it in Admin Console → Agent identity, and controls its audience. A builder can then attach the credential to an autonomous agent.

Learn more

Agent Identity [ROAD-1542]

Query Databricks Genie from Glean Assistant

Assistant can now route natural-language data questions to Databricks Genie spaces, allowing business users to ask questions about governed Databricks data without leaving Glean. Glean uses indexed Genie space metadata, such as titles, descriptions, sample questions, and table information, to select the best matching space. Direct Databricks SQL remains available as an explicit fallback when requested or when no suitable Genie space is found. Results are returned in Assistant while Databricks permissions continue to control what each user can access.

How to access: Your admin must configure the Databricks connector and tools, verify that Genie spaces are indexed, and enable the Databricks tools for Chat under Admin Console → Platform → Tools. Then ask Assistant a question about your Databricks data.

Learn more

Databricks in Assistant

Personalize Glean Home with optional interactive pets

Glean Home can display an optional interactive pet that stays on the page, provides lightweight chat and meeting alerts, and lets you run chats in the background while you continue working. You can hide or snooze the pet, switch between pets, and choose which notification types it can show. Some pets unlock as you use Glean Home, turning common actions such as searching, creating an artifact, or visiting People into optional discovery moments. The experience is controlled by your organization’s User Delight setting and may not be enabled everywhere.

How to access: When User Delight is enabled by your admin, open Glean Home and use the pet menu to choose, hide, snooze, or configure the pet and its notifications.

Glean Home [ROAD-1601]

Catch up on the last two minutes of a live meeting in one click

When you join a live meeting late or lose track of the discussion, Meeting Notes can now summarize the most recent part of the transcript for you. After the transcript has more than two minutes of content, select Catch me up above the composer to get a concise recap of what was just discussed. This helps you rejoin the conversation without asking others to repeat themselves and keeps you focused on the meeting instead of on note-taking. The summary is generated from the live transcript already available in the meeting.

How to access: Open a live meeting transcription in Glean desktop. After more than two minutes of transcript are available, select Catch me up above the composer. Meeting Notes access is controlled by your admin under Admin Console → Assistant → Meeting Notes.

Learn more

Meeting Notes [ROAD-769]

🤖 AI & Model Updates

A faster default model for Adaptive Mode, Fast Mode, and Search

Adaptive Mode, Fast Mode, and Search in Chat now use GPT-5.4-mini as their default model, delivering faster, more efficient responses in these modes.

How to access: No action needed — these modes use the new default model automatically.

Assistant

Set the default model for workflow agents

Admins can now configure the default model set that powers workflow agents, giving teams more control over which model is used for workflow-agent steps.

How to access: In the Admin Console, set the default model set for workflow agents in your model configuration.

Agents

Choose GPT-5.6 models across Assistant and Agents

GPT-5.6 Sol, GPT-5.6 Terra, and GPT-5.6 Luna are now available as premium models across supported Assistant and Agent surfaces. The model hub gives admins control over which models are enabled, while users can choose from the models made available to them when model selection is supported. Luna is initially focused on eligible agent workflows, while Sol and Terra are available in supported model-choice surfaces. Premium models may affect usage-based consumption, so administrators can manage availability according to their organization’s needs and billing policy.

How to access: In Admin Console → Platform → LLMs, enable the GPT-5.6 models in the Model Hub. Users can then select supported models from the Assistant model picker or Agent model settings.

Learn more

Models

Configure AI action models by deployment and workflow

Admins can now configure the model used for AI-powered actions with clearer deployment-aware controls. The configuration supports choosing an eligible model, setting visibility for the people who can use the action, and keeping the action aligned with the model availability of the deployment. This helps platform owners balance response quality, cost, and access while giving users a predictable experience when they invoke actions such as summarization, drafting, or other AI-assisted workflows.

How to access: Open the relevant action in Admin Console → Platform → Actions, select its configuration, and review the available model and visibility settings. The choices shown depend on your deployment and enabled models.

AI Actions

Use Claude 4.1 in Assistant

Claude 4.1 is now available for supported Assistant experiences, giving users another model option for tasks that benefit from Anthropic’s latest available capabilities. Administrators retain control over whether the model is enabled and which users or workflows can access it. The model appears alongside other configured models rather than changing every user’s existing selection automatically, so teams can evaluate it within their current model governance and usage-management practices.

How to access: An admin must enable Claude 4.1 in Admin Console → Platform → LLMs. Once enabled for your deployment, select it from the Assistant model picker where model choice is available.

Learn more

Assistant Models

Select models separately for individual applications

Model selection can now be configured at the application level instead of relying only on one broad deployment setting. This gives administrators more control over which applications can use particular models and helps align model access with the needs, risk profile, and cost expectations of each workflow. Users see only the models made available to the application and audience they are using, so the experience remains governed without requiring every application to expose the full model catalog.

How to access: Open Admin Console → Platform → LLMs or the relevant application’s model configuration and review the enabled models and visibility settings. Exact options depend on your deployment.

Learn more

Model Administration

New Glean Key autonomous agents start with GPT-5.6 Luna

New autonomous agents on Glean Key now use GPT-5.6 Luna as their default model, giving new agents a more capable starting point without requiring you to choose a model during setup. Existing autonomous agents keep their current model settings, so this change does not silently alter agents already in use. Workflow-agent defaults are also unchanged. For administrators, the Glean Key model controls now make the transition clearer: Luna can be enabled while it is off, and GPT-5.4 can be disabled after Luna is enabled. Legacy large and small workflow defaults are no longer presented as default choices for Glean Key.

How to access: Create a new autonomous agent to use the new default automatically. To review Glean Key model settings, open Admin Console → Models and select the Glean Key model configuration.

Learn more

Agents and Model Administration

⚠️ Critical Updates

Knowledge Profile searches now surface only company-wide pinned results

For searches performed with a Knowledge Profile, private and department-scoped pinned results are no longer shown as suggested results; only company-wide pins appear. This keeps pinned suggestions aligned with content that is broadly visible across your organization.

How to access: No action needed — this applies automatically. If you previously relied on private or department-scoped pins appearing under a Knowledge Profile, note that only company-wide pins will now surface.

Search

Restrict public links to people who already have access

Public-link settings now make it easier to prevent accidental broad sharing. Administrators can choose whether users may create public links and can restrict link creation to specific groups. When restricted sharing is selected, only people who already have access to the underlying content can open the link. This gives organizations a clearer control point for balancing collaboration with information protection and helps prevent a link from becoming an unintended access path to a chat, artifact, or other user-generated content.

How to access: Admins can review public-link controls in the organization’s sharing and access settings. End users will see the permitted sharing options when they open a chat or artifact’s share dialog.

Sharing and Permissions

Connect MCP hosts with a simpler OAuth consent experience

Connecting a third-party MCP host to Glean now uses a single, clearer user-facing permission instead of exposing a long technical list of OAuth scopes. The consent screen explains that MCP access operates within your existing permissions and the settings approved by your admin, while Glean narrows the underlying scopes to the resources the host needs. This makes the first connection easier to understand without removing the admin and per-tool controls that govern what the MCP host can access.

How to access: Admins who already use the Glean OAuth Authorization Server do not need to change anything. To enable it for a new deployment, open Admin Console → Settings → Third-party access (OAuth) and enable Glean OAuth Server. Users then connect their MCP host and approve the simplified consent screen.

Learn more

Glean MCP OAuth

Plan for the retirement of Claude Opus 4.1

Claude Opus 4.1 is being retired from Glean model availability. Administrators should review agents, actions, and other workflows that explicitly select this model and move them to a supported replacement before the retirement date. Existing workflows may continue to run during the transition window, but they will need a supported model after the model is removed. Reviewing model usage now helps you avoid unexpected changes to production agents and gives you time to test a replacement under your organization’s access and usage policies.

How to access: Open Admin Console → Platform → LLMs and review configurations that use Claude Opus 4.1. Choose a supported replacement and save the updated configuration before the retirement deadline shown in your deployment.

Learn more

Model Administration

Move workflows off GPT-5.1 before it is retired

GPT-5.1 is being removed from supported model choices. If an agent, action, or application explicitly depends on GPT-5.1, update that configuration to a supported model before the retirement date. The change is intended to reduce reliance on an older model version while preserving a clear migration path through the Model Hub. Administrators should also review any model-specific prompts, output expectations, or cost assumptions before switching production workflows to a replacement.

How to access: Open Admin Console → Platform → LLMs, search for GPT-5.1 in your enabled-model and workflow configurations, and select a supported replacement before the retirement deadline shown for your deployment.

Learn more

Model Administration

Control application and tool access with clearer admin settings

Administrators now have more consistent controls for managing who can use connected applications, tools, and agent capabilities. Access settings can be applied to users or groups, and the admin experience includes clearer guidance about where a setting takes effect and how it affects end-user visibility. These controls help teams roll out integrations safely, limit sensitive actions to the right audiences, and troubleshoot access without requiring users to guess whether a feature is unavailable or simply restricted by policy.

How to access: Open the relevant application, tool, or agent configuration in the Admin Console and review its Visibility, Access, or Configuration settings. Use the available user or group controls to define the intended audience.

Admin Access Controls

Admins can control who receives audio artifacts during rollout

Glean Key administrators now have explicit controls for audio artifacts. You can let Glean manage the rollout, enable audio artifacts for everyone, limit them to admins, or turn them off. These controls help you introduce podcast-style audio artifacts at a pace that matches your organization’s privacy, compliance, and adoption requirements. The setting applies to new audio generation; an audio artifact that has already finished generating remains playable even if the setting later changes. If access is revoked while generation is in progress, the generation is stopped cleanly rather than remaining stuck.

How to access: Open Admin Console → Glean Assistant → Settings → Audio artifacts on a Glean Key deployment, choose the audience setting, and save. Customer Key deployments do not show this setting.

Learn more

Audio Artifacts

Delegate MCP administration without granting full admin access

The MCP Server Moderator role is now enabled by default in deployments that do not provide an explicit override. Administrators can use this role to delegate Glean MCP server management and MCP Insights access without granting broader administrator permissions. The role remains governed by the deployment-level feature setting, so organizations that need a stricter permission model can disable it. This change makes it easier to separate day-to-day MCP administration from organization-wide administrative authority while keeping access controlled through existing role management.

How to access: Assign the MCP Server Moderator role through your deployment’s Admin Console role and permissions controls. If your deployment explicitly disables the role, it will remain unavailable.

MCP Server Administration

Departmental moderators now see the publish action they can actually use

Departmental agent moderators now receive an action that matches their publishing scope. When an agent was created outside the departments a moderator oversees, the builder previously displayed Publish, even though the publish request would be rejected. The action now appears as Request to publish, making the next step clear and avoiding a failed publish attempt. Moderators can still publish agents created within their departments, and organization-wide agent admins are unaffected. This change makes agent governance more predictable for teams that use department-scoped moderation.

How to access: Open an agent in the builder as a departmental moderator. The available action is shown in the agent’s publishing controls and reflects the creator’s department and your assigned scope.

Agent Permissions

Notifications work without customer-specific enablement

Notifications delivered through Glean’s Notify platform now work by default, removing the need for customer-specific enablement before supported notification workflows can operate. This reduces setup friction for organizations using notification delivery and makes behavior more consistent across deployments. There is no change required for customers who already use the platform. Existing notification configuration and delivery paths continue to apply; the improvement is that the underlying delivery capability is available by default instead of requiring an additional customer-level activation step.

How to access: No action is required. Use supported notification workflows as usual; delivery is available automatically through the Notify platform.

Notifications

Tableau connections work with older supported REST API versions

The Tableau connector now starts with REST API version 3.23 instead of 3.27. This improves compatibility for Tableau Server and Tableau Cloud environments whose maximum supported API version is below 3.27, reducing setup and crawl failures caused by a version mismatch. The change is backward-compatible for environments that support newer Tableau APIs, and it does not require you to change your existing Tableau configuration. Glean continues to use the connector’s supported setup and authentication flow while requesting an API version that works across a broader range of Tableau deployments.

How to access: No action is required. Use the Tableau connector setup and crawl workflows normally. Existing configurations continue to use the compatible default.

Tableau Connector

🌅 Deprecations & Sunsets

Gemini 3 Pro has been removed

Gemini 3 Pro has been removed following its March 25, 2026 deprecation date and is no longer available for selection in Assistant or Agents.

How to access: If your organization used Gemini 3 Pro, switch any affected Assistant or agent configurations to another available model in the Admin Console.

Models

🎨 Enhancements

Consistent meeting titles across chat, transcripts, Notes, and Summaries

Meeting titles now stay in sync across chat, transcripts, Notes, and Summaries after you rename a meeting or generate a Summary, so you see the same title everywhere.

How to access: No action needed — titles stay consistent automatically.

Meetings

Faster data refresh in Canvas artifacts

Refreshing data in a Canvas artifact is now faster and skips an unnecessary agent interaction when the artifact supports deterministic refresh.

How to access: No action needed — refresh is faster automatically for supported artifacts.

Artifacts

Spot chats that ran into an error from the left navigation

Conversations that hit an error before finishing now show an error indicator in the left navigation, making it easier to find chats that may need to be retried or reviewed.

How to access: No action needed — the indicator appears automatically next to affected chats.

Assistant

See read and write tool counts for external MCP servers at a glance

When choosing external MCP tools in the Admin Console, each server now shows how many read and write tools it provides before you expand it, and servers that provide no tools are no longer listed.

How to access: In the Admin Console, open external MCP tool selection to see per-server read and write tool counts.

MCP

Choose whether to notify teammates when sharing an agent

When you share an agent with new teammates, you can now decide whether to notify them. A "Notify teammates" checkbox appears whenever you add teammates, so you can alert them to their new access or share quietly.

How to access: When sharing an agent, use the "Notify teammates" checkbox to control whether new teammates receive a notification.

Agents

LumApps connector now captures deleted and archived content

The LumApps connector now supports webhooks for deleted and archived content, keeping search results more accurate as content changes in LumApps.

How to access: Available automatically on the LumApps connector; no additional setup is required.

Connectors

Clearer GitLab Cloud connector setup guidance

GitLab Cloud setup now guides admins to use either a Group Access Token or a Service Account token scoped to a single top-level group, with links to official GitLab documentation and clearer notes on when webhooks are created automatically versus configured manually.

How to access: In the Admin Console, follow the updated GitLab Cloud connector setup flow.

Connectors

Export historical user spend, including zero-usage and deleted users

Billing moderators can now export historical user spend, including users with zero usage and deleted users, for more complete cost reporting.

How to access: In the Admin Console billing area, billing moderators can export historical user spend.

Billing

Per-app and per-user spend limits with the AI gateway

When the AI gateway is enabled, monthly spend limits are now enforced per connected app and per user within each app, so a budget set on one app is evaluated independently of others. Deployments that do not use the AI gateway are unaffected and continue to enforce workspace-wide, per-user, and per-agent limits.

How to access: With the AI gateway enabled, set per-app and per-user monthly spend limits in the Admin Console.

Billing

Splunk MCP server is now connection-verified with a setup guide

The Splunk MCP server template now appears as "Connection verified" in the MCP catalog and ships with step-by-step instructions for connecting Splunk using a dedicated service account, a least-privilege role, and an encrypted MCP token.

How to access: In the MCP catalog, select the Splunk template and use "Show setup instructions" to complete service-account setup.

MCP

Support for the latest Model Context Protocol revision

Glean's MCP servers now support the 2026-07-28 revision of the Model Context Protocol specification. Clients that speak the newer revision negotiate it automatically, and clients on earlier revisions continue to work unchanged.

How to access: No action needed — supported clients negotiate the new revision automatically.

MCP

More resilient Slack tools under heavy concurrent use

Slack tools used by Assistant, voice, and agents are now more resilient when many requests run at once. Glean no longer opens a new Slack connection for every skill lookup and tool dispatch, so Slack per-account connection rate limits no longer reject requests mid-turn and cause an assistant or scheduled agent to abandon its remaining steps.

How to access: No action needed — this reliability improvement applies automatically.

Slack

Zendesk Extension Tools action pack is now available to set up

Resolved an issue where the Zendesk Extension Tools action pack did not appear in the list of supported action packs and therefore could not be configured. It is now available to set up.

How to access: In the Admin Console under Actions, add the Zendesk Extension Tools action pack.

Actions

Filter Glean Home suggestions by assistant capability

Glean Home suggestions can now be filtered by the assistant capabilities they use. This gives admins and product teams a more focused way to evaluate or manage suggestion experiences when a deployment exposes only selected workflows. The filter separates suggestions based on the underlying capability rather than requiring you to inspect each card individually, making it easier to understand what types of actions are being promoted and to keep the Home experience aligned with the workflows available to your organization.

How to access: Open the Glean Home suggestion management or configuration view and use the assistant-capability filter when reviewing available suggestions. The filter appears automatically where suggestion controls are enabled.

Glean Home

Review and export expense data from Glean Expenses

Glean Expenses now supports a clearer review workflow for expense information, including improved access to the relevant expense records and exportable data. This helps finance and operations teams inspect expense activity in one place and move the information into their existing reporting processes when needed. The update focuses on making expense data easier to retrieve and use without changing the underlying expense source or approval process.

How to access: Open Glean Expenses from the application area available to your organization, review the expense records, and use the available export action when you need a report.

Glean Expenses

Upload more text-based file formats safely

File uploads now support more than 50 additional text-based formats, including TSV, TOML, RST, and Markdown. Glean continues to block unsafe or unsupported files such as executables, archives, scripts, and audio or video formats where applicable. This makes it easier to bring structured text and configuration files into Assistant for analysis, summarization, and content generation without weakening the existing security checks that validate file types before processing.

How to access: In Assistant, choose the file-upload control and select a supported text file. If your organization has file upload disabled, an admin can enable it from the Glean section of workspace Settings.

Learn more

File Upload

Keep working with AI in full-screen artifacts

Full-screen artifacts now include a floating composer so you can ask for edits, answer clarifying questions, and continue iterating without returning to the side-by-side chat layout. The composer opens by default the first time you view an artifact in full screen. If you prefer an uncluttered view, minimize it; that choice stays in effect for the current artifact editing session. This is especially useful for dashboards, interactive pages, slide decks, and MCP apps where the full-screen view is the main working surface.

How to access: Open an artifact and select full-screen view. Use the floating composer to ask for changes or questions, or minimize it when you want to focus on the artifact.

Learn more

Artifacts and Canvas

Read Assistant more comfortably with updated typography

Assistant now uses Inter typography with updated spacing and visual hierarchy across the core chat experience. Responses, controls, and structured content have a more consistent appearance across operating systems, making dense workflows easier to scan and giving the product a more predictable visual rhythm. The update does not change how you start chats, use tools, or access existing content; it improves the readability and polish of the surfaces you already use.

How to access: No action is required. The updated typography and spacing appear automatically in Assistant.

Assistant

Open Assistant with the new application URL

Assistant now has a dedicated application URL, making it easier to bookmark, link to, and open the Assistant experience directly. The new URL supports a clearer entry point for users who want to move straight into chat rather than starting from a general Glean surface. Existing navigation continues to work, so teams can adopt the direct URL gradually without changing links or workflows immediately.

How to access: Open the Assistant application from your existing Glean navigation or use the direct Assistant URL provided for your deployment.

Assistant

Use more skills and configure their visibility

Skills now support additional setup and access controls so organizations can make reusable agent capabilities available to the right audiences. Admins and builders can configure skills as part of their agent workflows, while visibility settings help prevent a skill from appearing to users who should not use it. This supports more consistent reuse of approved instructions and capabilities across agents without requiring each builder to recreate the same behavior.

How to access: Open the Skills area in Glean or the relevant agent builder, then review the skill’s setup and access controls. Your admin may need to enable or grant access to the skill.

Skills

Show more connected sources and actions in the Assistant app

The Assistant app now exposes additional connected sources and actions from the application experience, making it easier to discover what Glean can use to answer questions or complete work. The expanded surface helps users find supported integrations without memorizing connector names or navigating through separate configuration pages. Availability still depends on the connectors and actions enabled for your organization, so users see a governed set of options aligned with their deployment.

How to access: Open the Assistant app and use its sources or actions controls to review what is available. If an option is missing, ask an admin to verify connector and action visibility settings.

Assistant App

Export and review uploaded files with clearer format support

The file-upload experience now better communicates which text-based formats can be selected and processed. This complements the expanded format support by reducing confusion when a valid file does not use one of the most common extensions. Glean validates files after selection and continues to apply server-side security checks, so the broader format support improves compatibility without treating every file type as acceptable. You can upload supported files to analyze them, ask questions, or create content from their contents.

How to access: In Assistant, select the upload control and choose a supported file. If a file is rejected, check its format against the supported file types in the public file-upload guidance.

Learn more

File Upload

See generated image batches more naturally on native mobile

When Glean generates several images in native mobile chat, the preview now uses one large hero image with supporting thumbnails instead of desktop-style artifact framing. The whole preview is a single accessible tap target, so you can open the full image canvas without hunting for a small control. The layout is designed for mobile screens and keeps the available images visible at a glance. Desktop and mobile web previews are unchanged, so this improvement is focused on making generated image results easier to scan and open in the native mobile experience.

How to access: In the Glean mobile app, generate or open a chat response containing multiple images. Tap the hero image or thumbnail area to open the full image canvas.

Mobile Chat

Find artifact types faster with clearer filter names and icons

Search filters now use friendlier names and an icon that matches each artifact type. In artifact-focused searches, documents, interactive artifacts, meeting notes, images, and other supported types are easier to distinguish than when every row used the same generic artifact icon. The With artifact filter also uses clearer wording in Assistant search. These changes reduce the time needed to understand what a filter represents and make it easier to narrow results when you are looking for a specific kind of artifact.

How to access: Open Search, optionally scope to Assistant or Artifacts, and open the With artifact or Type filter. The updated labels and icons appear in the filter list automatically.

Search and Artifacts

Review all agent versions in one timeline

The agent builder’s version picker now presents drafts, saved versions, staged changes, and published versions in one time-ordered timeline instead of splitting them between Drafts and Published tabs. This gives you a single view of an agent’s history and makes it easier to understand which version is current, which changes are ready to publish, and what was saved previously. Existing version actions remain available, including selecting a version and managing staged or saved versions where those actions apply.

How to access: Open an agent in the builder and open the version picker. Review the timeline from the current version through earlier drafts, saved versions, and published checkpoints.

Agent Builder

Search, sort, and evaluate agents more easily when configuring an MCP server

Admins configuring a Glean MCP server can now use a more capable agent picker. Search and familiar filters help you find the right agents, sorting and usage metadata provide more context, and descriptions are available without leaving the picker. Agents that cannot be selected, including disabled human-in-the-loop agents, remain visible with an explanation rather than disappearing and leaving you unsure why they are unavailable. The staged selection model remains in place, so your changes are applied when you save the MCP server configuration.

How to access: Open Admin Console → Glean MCP servers, edit an MCP server, and open its Agents section. Use the search, filters, sorting controls, and agent details in the picker before selecting Save.

Glean MCP Server

Review every configured GitHub extension in one place

The GitHub connector setup page now shows all configured file extensions used for indexing instead of limiting the visible list to a small fixed number. When many extensions are configured, they appear in a scrollable area so you can review the full configuration without losing access to the search and add controls. This makes it easier to verify that the connector covers the file types your engineering and documentation teams rely on and reduces the risk of overlooking an extension that was configured earlier.

How to access: Open the GitHub connector’s setup page in the Admin Console and review the configured file extensions. Scroll within the extension list to see entries beyond the initially visible rows.

GitHub Connector

Usage tables now label model spend by the selected period

The Models tables in the Admin Console now use the same period-aware labels as the users and departments tables. For the current month, the column is labeled MTD usage; for a previous month or all-time view, it is labeled Usage. The same wording is used in relevant application breakdowns and CSV exports. This makes it easier to interpret the numbers you are reviewing and avoids confusing a month-to-date value with a historical or all-time total.

How to access: Open Admin Console → Usage → By models and choose the current month, a previous month, or All time. The table and CSV export header update to match the selected period.

Usage and Billing

Open directory details without leaving AI Security Insights

The AI Security Insights user list now supports directory person cards. When you need more context about a high-risk user, you can hover over the user in the list and view the available directory details without navigating away from the insights dashboard. The existing user, department, pagination, and risk information remain in place, so the person card adds context without disrupting the review workflow.

How to access: Open /admin/agent-guardrails/insights and hover over a user in the user issues panel to open their directory person card.

AI Security Insights

MCP server lists now count inherited managed tools correctly

MCP server lists now include tools inherited from Glean-managed toolsets when showing the enabled tool count. Previously, a server could show zero tools in the list even though its detail page contained inherited tools. The count now reflects the union of enabled built-in tools and available managed-toolset members without double-counting overlaps or including unavailable tools. This gives admins a more accurate at-a-glance view when comparing servers and deciding which configuration to open.

How to access: Open the MCP server list in the Admin Console. The corrected tool count appears automatically for servers that use Glean-managed toolsets.

MCP Server Administration

Sharing language now makes restricted access explicit

The restricted-access option in sharing dialogs now says Restricted to only people with access. The wording clarifies that choosing this option does not grant access to new people; it limits access to people who already have it. Sharing behavior and the dialog layout are unchanged. The clearer label helps you choose the right sharing scope, especially when you are sharing a chat or artifact with a group and need to understand whether recipients must already be permitted to view it.

How to access: Open a chat or artifact sharing dialog and select the restricted-access option. The updated explanation appears in the sharing controls.

Sharing and Permissions

Assistant Replies preferences stay with each chat

Your Assistant Replies preference is now saved separately for each chat. If you turn Assistant Replies off in one conversation, leave it, and reopen it later, that chat remembers your choice. A different chat can keep a different setting, so changing the preference in one conversation does not change it elsewhere. The preference is stored locally in the web experience and does not modify the backend chat data, preserving the existing chat history and collaboration behavior.

How to access: Open any chat, change the Assistant Replies preference, and return to the conversation later. The setting is restored for that chat automatically.

Assistant Chat

Connect Zendesk MCP with service-account credentials

Admins can now configure a Zendesk MCP server using a confidential OAuth client’s service credentials instead of relying on a user’s Zendesk connection. Enter the Zendesk subdomain and the client credentials in the service-account template; Glean exchanges them for short-lived access tokens and exposes the existing Zendesk read tools to agents that use the server. This supports service-identity workflows while keeping the existing per-user Zendesk OAuth path available for customers who need it.

How to access: In the Admin Console, add the Zendesk MCP (Service Account) template, enter the Zendesk subdomain, and provide the confidential OAuth client credentials in the credential field. The service-account template may need to be enabled for your deployment.

Zendesk MCP

Customer log streams include LLM response latency metrics

Customer event streams now include per-call time-to-first-token and total latency for LLM calls. Time-to-first-token measures how long it takes before the model begins responding, while total latency captures the full duration of the call. These fields give administrators and platform teams more context when monitoring AI responsiveness, investigating slow experiences, or comparing performance across usage patterns. The change adds observability to customer-visible LLM call events without requiring application changes or changes to the way users submit Assistant or agent requests.

How to access: No action is required. Review LLM call events in your existing customer log stream; the latency fields are included automatically for new events.

Customer Event Logs

🐛 Fixes

Reliable indexing of OneDrive and SharePoint updates

Fixed a rare crawling issue in OneDrive and SharePoint where a failure between fetching an update and storing it could result in a permanently missed update. Updates are now stored before the crawl position advances, so changes are no longer lost.

How to access: No action needed — the fix applies automatically.

Connectors

OneDrive documents that previously failed to index are now processed correctly

Fixed an issue that prevented some OneDrive documents from being indexed. OneDrive documents are now processed correctly, with SharePoint handling unchanged.

How to access: No action needed — the fix applies automatically.

Connectors

Glean Chat recovers from errors on follow-up turns

Fixed an issue where Glean Chat could fail with a "Something went wrong" error on follow-up turns when the underlying AI provider could not decrypt a prior reasoning context. Glean now clears the invalid context and recovers the conversation turn.

How to access: No action needed — the fix applies automatically.

Assistant

Zoom recordings open directly from search results

Zoom meeting recording search results now link directly to the recording's share URL with the play passcode included, so the recording opens without a separate passcode prompt. Links to the older recording details page continue to resolve to the same result.

How to access: No action needed — recording links open directly.

Zoom

Slack direct messages reliably appear in federated search

Fixed an issue where Slack direct messages found through federated search could be missing from results on deployments that index Slack but exclude private messages from native search.

How to access: No action needed — the fix applies automatically.

Slack

Prevent connector reindexing from leaving stale or broken results

Connector reindexing could leave affected customer content in an inconsistent state after source changes, including results that pointed to missing pages or returned 404 errors. The reindexing flow now handles those changes more safely so refreshed content is available from the connector without leaving stale references behind. This improves trust in search results after large updates, deletions, or connector maintenance and reduces the need for customers to retry indexing or investigate broken links manually.

Connectors and Search

Avoid request failures after a usage limit is reached

Some customers could receive HTTP 429 errors on every response after reaching a configured usage limit, even when the request should have returned a clear limit message or allowed a non-billable experience to continue. The usage-limit handling path now returns the correct response behavior instead of repeatedly surfacing a generic rate-limit failure. This makes limits easier to understand for end users and gives administrators a more predictable control mechanism when a user, agent, or organization reaches its configured threshold.

Usage Limits

Keep Deep Research jobs from running out of memory

Large or complex Deep Research jobs could consume too much memory and fail before producing results. The workflow now handles the research execution more efficiently so longer investigations are less likely to terminate because of an out-of-memory condition. This helps users continue using Deep Research for substantial questions while preserving the context and intermediate work needed to produce a useful response instead of forcing them to restart with a smaller request.

Deep Research

Keep AWS connector links working after deployment changes

Some AWS-hosted connector experiences could generate links that pointed to an unavailable endpoint, resulting in 404 errors when users opened a connector result or management path. The URL handling now uses the correct AWS deployment context, so connector links resolve to the expected destination. This restores normal navigation for affected AWS-hosted customers and reduces the need to repair links or retry connector actions manually.

AWS Connectors

Apply connector permissions consistently through hierarchy changes

Connector permissions could be applied incorrectly when access was inherited through parent and child connector relationships. The permission evaluation path now preserves the intended hierarchy so users receive the correct access to indexed content, including after connector configuration changes. This reduces the risk of users seeing content they should not access or losing access that an administrator intended to grant, and it makes connector governance more predictable for organizations with layered source configurations.

Connector Permissions

Long Deep Research chats are less likely to fail before answering

Deep Research sessions with very long histories could fail with an error instead of returning an answer. The route now handles large conversation histories without sending the entire history inline in the request, allowing the session to continue while preserving earlier context. This is especially helpful when you return to an extended research thread and need to build on prior findings rather than start over.

Deep Research

GitHub code search results now open the correct file

GitHub federated code search results could lead to 404 pages when the result URL used an invalid reference. Results now resolve through the repository’s default branch, so selecting a file opens the expected GitHub page instead of an error. This restores the basic workflow of searching code in Glean and jumping directly to the matching file for inspection.

GitHub Search

New Slack Real Time Search connections can complete authorization

New Slack Real Time Search connector setups could fail before OAuth began because the authorization URL was not generated. The setup flow now handles newly created instances correctly, allowing the authorization URL to be created and the OAuth process to start. Existing Slack Real Time Search instances continue to use their saved MCP scope settings.

Slack Real Time Search

Long-running Spaces chats can continue past large-history limits

Long-running Spaces conversations could fail when their history became too large for the request sent to the AI workflow. The conversation history is now read directly from stored chat events for this route, avoiding the request-size limit while preserving the context needed for follow-up turns. This helps you continue an established Spaces discussion instead of losing the thread when the conversation grows.

Spaces

Tools work correctly across multiple Microsoft 365 instances

When multiple Microsoft 365 instances were connected, tool selection could use the wrong identity mapping, causing tools to be missing or unavailable in some contexts. Tool selection now respects the identity mapping for additional Microsoft 365 instances, so the tools exposed to an agent match the configured connection. This improves reliability for organizations that separate Microsoft 365 environments or maintain more than one connected instance.

Microsoft 365 Tools

Original source

User Features

Model Selector in Builder Assistant

Builder Assistant now includes a model selector so agent builders can choose which large language model powers each compiler chat. The chat bar shows the current model, and every response clearly indicates which model was used, matching the Assistant experience. This gives admins more control over cost and quality when Builder Assistant usage is billed, while keeping the default "Auto" model for new sessions.

How to access: Administrators must enable for your organization. Builder Assistant with model selection is available to all customers who can access the Agent Builder and use Builder Assistant. Admins should ensure that any premium models they want builders to use are enabled in their model settings. No additional configuration is required beyond existing model and billing setup. This feature may be subject to usage-based pricing. Learn more in Glean's Help Center.

Learn more

Agents [ROAD-1566]

Call Third-Party Agents from Glean Agents via A2A

Glean agents can now call external third-party agents that support the A2A (Agent-to-Agent) protocol, so complex workflows can seamlessly span Glean and other agent platforms. Admins register external A2A agent servers in the Admin Console with OAuth credentials, and agent builders can add those external agents as sub-agents in Plan & Execute and Auto Mode workflows. This lets Glean agents delegate specialized tasks to domain-specific agents built on any A2A-compliant platform, while preserving per-user authentication, permissions, and governance across systems. This feature is available for all customers using Glean Agents and may be subject to usage-based pricing.

How to access: Administrators must enable for your organization. To enable this capability, a Glean admin must go to the Admin Console and register one or more external A2A agent servers, including their A2A agent card URL and OAuth client credentials, under Settings → Third-party agents. Admins should then configure which external agents are available to use in agent workflows via Settings → Assistant → Agent orchestrator. Once configured, agent builders can select and add these third-party agents as sub-agents in the agent builder; no additional setup is required for end users beyond any standard sign-in or consent flow enforced by the external platform. This feature may be subject to usage-based pricing. Learn more in Glean's Help Center.

Learn more

Agents [ROAD-1560]

Adaptive Reasoning [Customer key]

Adaptive reasoning dynamically decides how long to reason based on the user's query. This makes Glean Assistant feel faster for easy queries while preserving its quality for more difficult ones.

Learn more

Assistant [ROAD-1376]

Glean Assistant can now query BigQuery with natural language or SQL

Glean Assistant can now query BigQuery directly through BigQuery MCP, Google's official MCP server for BigQuery, so users can ask questions in natural language or run SQL against approved projects and datasets without leaving Assistant. This makes Glean a front door to your warehouse so business users get governed, self-serve access to live data while data teams stay in control of what is exposed and how it is queried.

Learn more

Connectors [ROAD-1165]

Claude Opus 5 in Assistant and Agents

Claude Opus 5, Anthropic's latest premium Opus-class model, is now available in Glean Assistant model choice and in Agents. It delivers Anthropic's highest-quality performance for coding, agentic workflows, and long-horizon enterprise tasks, giving teams a powerful new option when they need stronger reasoning than standard models can provide. Customers already using Opus-class models can adopt Opus 5 as their new Anthropic premium choice without changing how they work in Assistant or the agent builder. This feature may be subject to usage-based pricing; see your plan details for how premium models are billed.

How to access: Administrators must enable Claude Opus 5 for your organization before users and agent builders can select it. In the Admin Console, go to Platform → LLMs (Model Hub), make sure your Anthropic provider configuration is set up, and turn on the Claude Opus 5 model for Assistant and Agents. If you want to limit access, you can hide premium models from model pickers or restrict Claude Opus 5 to specific departments using the existing model governance controls. Once enabled, end users can pick Claude Opus 5 from the model picker in Assistant, and builders can select it in the agent builder for eligible agents and steps. This feature may be subject to usage-based pricing. Learn more in Glean's Help Center.

Learn more

Agents [ROAD-1592]

Gemini 3.6 Flash in Assistant and Agents

Gemini 3.6 Flash is now available in Glean Assistant model choice and in the Agents model hub. Gemini 3.6 Flash offers significantly better cost efficiency than larger frontier models while maintaining strong reasoning for well-scoped agent tasks.

How to access: Glean Universal Model Key: Gemini 3.6 Flash appears automatically in the Assistant model choice dropdown and Agents model hub. Admins can exclude it via Admin Console → Platform → Models.

Customer Key: Gemini 3.6 Flash is available via Admin Console → Platform → Models. This feature may be subject to usage-based pricing. Learn more in Glean's Help Center.

Learn more

Agents [ROAD-1584]

A simpler setup flow for end users connecting to Glean MCP

We've refined the end-user setup experience so the default Glean MCP server is the clearest path for most users, while still preserving access to additional servers for advanced use cases. The updated flow also makes connected apps and authorization wording easier to understand, and gives users clearer setup options depending on the host.

How to access: End users can go to Settings → Third party apps and MCP and use the Glean MCP setup flow to pick a host and connect to Glean. The default server is highlighted as the recommended path, while additional servers remain available for advanced use cases.

Learn more

Assistant [PACT-331]

Assistant typography refresh and visual polish

Assistant now uses Inter typography with updated spacing and visual hierarchy for a consistent, easier-to-read experience across all devices.

Learn more

Assistant [ROAD-1526]

Glean Assistant on external AI platforms via A2A (Cross-platform)

Introduces a cross-platform A2A foundation that lets Glean Assistant run as a governed, first-class agent inside external AI platforms and agent runtimes, starting with Gemini and expanding to other ecosystems over time. The shared protocol and packaging model reuse Glean's enterprise context, permissions, and guardrails so users can ask Glean from their preferred AI surface without duplicating configuration or exposing extra data. Admins get a consistent way to authorize external platforms, manage which Glean capabilities are available where, and monitor usage across environments.

How to access: From the Glean admin console, enable external AI platform access/A2A and connect each supported platform using its service account or OAuth configuration. Use the A2A settings to choose which Glean agents (including Glean Assistant) can be invoked from each external platform and which user groups are allowed to use them. Refer to the Help Center setup guide for per-platform steps (for example, registering Glean Assistant in Gemini or other A2A-capable marketplaces). This feature may be subject to usage-based pricing. Learn more in Glean's Help Center.

Learn more

Connectors [ROAD-1463]

Simplified MCP OAuth consent and scope surface

Connecting third-party MCP hosts to Glean now uses a single, clear consent screen instead of a long list of technical OAuth scopes. The new experience groups MCP access into one user-facing permission, aligns that approval with each person's existing permissions and admin-approved settings, and narrows the underlying scopes to only the resources the host needs. Additional instrumentation on consent approvals and declines helps admins monitor rollout safety and refine their MCP strategy over time.

How to access: Admins already using the Glean OAuth Authorization Server for remote MCP servers do not need to take any action; users will automatically see the updated consent screen the next time they connect a host. To start using OAuth with remote MCP servers, open the Admin Console, go to Settings → Third-party access (OAuth), and enable the Glean OAuth Server. Then configure the MCP hosts you want to allow, and users can initiate the connection from their host and complete the simplified consent flow in their browser.

Learn more

Connectors [ROAD-1448]

NetSuite MCP support

NetSuite MCP support lets teams use Glean Assistant and agents to run live, permission-aware workflows against NetSuite on top of the existing indexed NetSuite connector. Users can ask Glean to look up customers, invoices, purchase orders, and other NetSuite records in natural language and pull the latest data into their conversations without switching tools. This capability is initially available to a small set of customers while we validate quality and coverage across common NetSuite use cases.

How to access: Admins can configure NetSuite MCP from the Glean Admin Console under Platform → Actions by adding the NetSuite MCP server as an action pack. During setup, provide the NetSuite MCP server URL and OAuth details so Glean can connect as an MCP host. Once the action pack is connected and its tools are enabled, eligible users can invoke NetSuite MCP actions from Glean Assistant and agents, with access governed by their NetSuite permissions.

Learn more

Connectors [ROAD-1336]

Get trusted Glean answers directly in Slack channels

Glean can now share trusted answers directly in Slack channels.

Glean can now post high-quality answers directly into Slack channel threads, so everyone can see them immediately without first clicking "view response." These public replies are generated only from content that's broadly visible in your company (such as public Slack channels and org-wide documents), so they're safe to share and easy to trust. Teammates can still personalize the answer with their own private context when they need a more tailored response.

How to access: To use public answers in Slack, first set up the Slack Real Time Search connector from the Glean Admin Console and install the Glean app in your Slack workspace or Enterprise Grid. If you have already done this, you can jump to the 'Embedded Integrations' tab to view the Glean in Slack setup page.

On this setup page, enable Gleanbot to respond in channels publicly and/or proactively. You can enable it for all public channels by default, or choose specific channels where you want Glean to participate.

Once enabled, channel members can ask questions in Slack and Glean will search across your company's public corpus (including public Slack channels) to generate shared responses, while still respecting each user's permissions for any follow-up personalization.

Learn more

Embedded Integrations [ROAD-908]

Persistent extension install button on Glean Home

Users who are actively using Glean but have not yet installed the browser extension will now see a persistent "Finish setup" style button on Glean Home that guides them to install the extension. This makes it clear that setup is incomplete and gives people a simple, always-available way to add the extension and unlock features like sidebar search, new tab page, and integrated search in apps like Google Drive and Box. The prompt appears only for eligible users and does not affect usage or billing.

Learn more

Glean Extension [ROAD-1581]

Admin Features

Admin dashboard now offers personalized recommendations of connectors, tools, and features to enable

The admin dashboard has been updated to include recommended connectors, tools, and features, alongside the existing notifications widget. The goal is to help admins discover the features that will open up new use cases for their users.

Learn more

Admin Capabilities [ROAD-1549]

Greenhouse connector migration to Harvest v3 APIs

Greenhouse connector Harvest v3 migration lets admins update their existing Greenhouse connector from legacy Harvest API credentials to Harvest v3 (OAuth) ahead of Greenhouse's August 31, 2026 deprecation deadline. Admins can add the new client ID and client secret in the Glean admin console without re-indexing or downtime, preserving existing search results and permissions. In beta: opt in from the admin console.

How to access: In the Glean admin console, open Connectors > Greenhouse > Setup and add Harvest v3 (OAuth) credentials (client ID and client secret). Keep the existing Harvest API Key and Webhook Secret in place, then save and wait for validation to complete.

Learn more

Connectors [ROAD-1393]

Granular controls for who can manage MCP servers

Admins can now delegate MCP server administration without giving full admin access. The new MCP Server Moderator role lets designated users create and manage MCP servers and view MCP Insights, making it easier to put ownership closer to the teams actually running AI workflows.

How to access: A Glean administrator can assign the MCP Server Moderator role to designated users so they can manage MCP servers and view MCP Insights without receiving broader admin permissions. Learn more in Glean's Help Center.

Learn more

Assistant [PACT-345]

More precise access control for MCP tools

Admins can now configure tool-level access controls directly from the Glean MCP server experience. This makes it easier to control which tools different users can access, while keeping permissions aligned across the broader tools and admin surfaces.

How to access: In the Glean Admin Console, open a Glean MCP server, click Add tools, go to External Tools, and open Tool advanced settings for a tool pack. From there, admins can set access for individual read and write tools, and those changes stay synchronized with the Tools admin experience. Learn more in Glean's Help Center.

Learn more

Assistant [PACT-344]

Glean is now available through the ChatGPT app marketplace

Glean's remote MCP server can now be enabled through the ChatGPT app marketplace, making it easier for admins to roll out Glean in ChatGPT and connect users to enterprise search and other Glean-powered workflows from that surface.

How to access: ChatGPT workspace admins can add the Glean MCP server from the ChatGPT admin connector experience and authenticate it with Glean OAuth. Once configured, users can enable Glean in individual ChatGPT conversations from the tools menu when they want to use it. Learn more in Glean's Help Center.

Learn more

Assistant [PACT-14]

Original source

User features

GPT-5.6 Sol, Terra, and Luna in Assistant and agents

OpenAI GPT-5.6 Sol, GPT-5.6 Terra, and GPT-5.6 Luna are now available in Glean Assistant and agents as premium models.

How to access: Administrators must enable this for your organization. This feature may be subject to usage-based pricing. Learn more in Glean's Help Center.

Learn more about premium models in Assistant

Agents [ROAD-1562]

Mix and match existing slides to create a new presentation with Glean

You can now build new presentations by mixing and matching slides from decks you already have in Glean, instead of starting from a blank prompt. Choose one or more source presentations, then let Assistant suggest which slides to keep, drop, or reorder based on the story you want to tell, while generating only the missing title, transition, and summary slides needed to make the deck feel cohesive. The final presentation is exported through the existing slide generation pipeline to PowerPoint or Google Slides, preserving your company's branding and template settings.

How to access: Administrators must enable this for your organization. This feature may be subject to usage-based pricing. Learn more in Glean's Help Center.

Learn more about mixing and matching slides

Assistant [ROAD-1537]

Trigger agents by events

Automatically run Glean Agents in the background when content changes across your connected data sources — Gong, Jira, Salesforce, Google Calendar, email, and more — no manual intervention required. Use triggers to scale repetitive processes, ensure timely actions, and keep work moving: agents can send Slack notifications, generate summaries and kick off downstream workflows, all within your organization's existing guardrails and permissions. It is Generally Available and enabled by default for all users. Admins can scope access from the admin console and check out the Glean Docs for supported sources and setup.

This feature may be subject to usage-based pricing. Learn more in Glean's Help Center.

Learn more

Agents [ROAD-731]

Interactive artifacts are now generally available

With interactive artifacts, users can create visual proposals, rich analysis, and refreshable dashboards. Each artifact lives natively in Glean with a stable URL, shows up in Glean search, and can be shared at private or org-wide visibility without exposing the underlying chat.

This release moves interactive artifacts out of preview and includes several improvements:

• Start from templates. Ready-made templates surfaced under the Create artifacts button in the library to help you spin up artifacts for common workflows.
• Style your artifact. Change the look and feel of an artifact so it matches your team's or company's visual style.
• See source citations. Artifacts now surface citations to the underlying sources, so you can trust the data being presented.
• Higher-quality output. When in auto mode, interactive artifacts will be generated by a model that is optimized for creating polished visualizations, such as Claude Opus 4.8 (if enabled in your deployment).
• Improved triggering. Glean is better at recognizing when a request is a good fit for an interactive artifact, so more users discover the feature naturally.

How to access: Interactive artifacts are available automatically in Assistant when the Interactive artifact setting is turned on in your Glean admin console. If Interactive artifact is turned off, Assistant will respond with standard text-only answers.

This feature may be subject to usage-based pricing. Learn more in Glean's Help Center.

Learn more

Assistant [ROAD-1540]

iOS shortcuts on home screen and lock screen

Launch a Glean Chat or Voice chat directly from your iPhone home screen or lock screen — one tap and you're live. Add the Glean Chat or Glean Voice widget, and on iOS 18, use the new Lock Screen and Control Center controls for even faster access. No app navigation required.

Learn more

Assistant [ROAD-1539]

Full-screen and navigation buttons on Glean-generated slides

Present Glean-generated slides directly inside Glean with built-in full-screen and navigation controls.

Learn more

Assistant [ROAD-1538]

Queue a follow-up chat during long-running Assistant chats

Queue your next prompt while Glean Assistant is still working. If a chat is already in progress, you can add a follow-up message to run automatically as soon as the current response finishes, so you can keep moving without interrupting the task or coming back later to retype it. Queued follow-ups also carry over composer settings, including file attachments and model selection. You can also review, edit, or remove queued follow-ups before they run.

Learn more

Assistant [ROAD-1507]

Render charts, diagrams, and visualizations in Assistant responses

When you ask for a chart or a diagram, Glean can render it as a compact, interactive widget right inside the chat response. You can then copy the widget's output as an image to drop into a doc, slide, or message.

How to access: Inline artifacts are available automatically in Assistant when the Inline artifact setting is turned on in your Glean admin console. If Inline artifact is turned off, Assistant will respond with standard text-only answers.

Learn more

Assistant [ROAD-1468]

Create interactive artifacts from templates

Templates help end users quickly create interactive artifacts for common workflows. Users can start from ready-made templates surfaced in the library under the "Create" button, then customize the generated artifact to fit their work.

This feature may be subject to usage-based pricing. Learn more in Glean's Help Center.

Learn more

Assistant [ROAD-1344]

Share artifacts with your entire organization via the share button

Artifacts can now be shared with your entire organization, making them discoverable via the library and search. This is in addition to existing sharing options like specific people or departments and anyone with the link, giving teams a clearer way to control how broadly their content is published.

Learn more

Assistant [ROAD-1331]

Group your chats, sources, and artifacts with Glean Projects

Glean Projects help users organize chats and content in Glean so related work is easier to group, revisit, and share.

How it works: Create a folder to keep relevant chats and artifacts together in one place. folders are private by default and can be shared when needed. Existing Collections will automatically migrate to folders, with existing permissions preserved.

Assistant [ROAD-909]

Writing profile removed from Personalization

The writing profile section has been removed from Settings → Personalization. A more detailed memory experience - with improved ways to view and manage your memories, writing style, and more - is coming soon.

Assistant

More streamlined chat conversation search

Use the new chat history search dialog to quickly find past Glean chats and agent runs from the left sidebar. Search by thread title keywords and filter between Assistant chats and specific agents so you can jump back into ongoing work without scrolling. Open chat history search from anywhere in Glean with the Cmd+K keyboard shortcut.

How to access: Admins: In the admin console, go to Platform → Assistant → Settings and, under Chat history, select a retention period (for example 90 days or 1 year) instead of Off, then save. End users: open Glean, use the Chats panel in the left sidebar, and either type in the chat history search bar or press Cmd+K to open the chat search dialog.

Learn more

Glean Home [ROAD-1499]

Admin features

Admin Capabilities [ROAD-1531]

Model governance by department for LLMs (Glean key and Customer key)

Admins can now configure model governance policies at the department level for LLMs, giving organizations more granular control over which models are available to different teams.

Learn more

Admin Capabilities [ROAD-1339]

Per-user usage visibility in Billing Console

Admins can now see AI and model usage at the individual user level in the Billing Console, not just as aggregate workspace totals. The per-user view helps billing and finance teams explain where spend is coming from, spot outliers or power users early, and export user-level usage to support internal chargeback or governance workflows. This update adds richer visibility only and does not change how limits or usage controls are enforced.

How to access: Admins and billing moderators can access per-user usage from the Admin Console by going to Usage and choosing the By users view. Access to these views is limited to roles with billing permissions. If you don't see Usage in the admin console, your Glean role may not include billing or admin access.

Learn more

Admin Capabilities [ROAD-1185]

Per-user, per-agent FlexCredit usage limits in admin console

Admins can now set per-user monthly usage limits directly in the Admin Console, with global defaults and per-user, per-agent overrides to match different roles and departments. When a user reaches their usage limit, Glean automatically pauses access to billable features. This gives customers more predictable AI costs, stronger governance controls, and a safer path to scaling usage.

How to access: From the admin console, go to Usage → Manage usage limits to configure cost guardrails.

Learn more

Admin Capabilities [ROAD-731]

Search and filter MCP servers in the Admin Console

Admins can now quickly find specific MCP servers in the admin console using a search bar on the Glean MCP server page.

How to access: From the Admin Console, go to Platform → Glean MCP server and use the Search by server name box above the server list to filter servers as you type. No additional configuration is required.

Learn more

Connectors [ROAD-1533]

Original source

User Features
JSM request routing and field mapping in tools
Glean tools now support end-to-end Jira Service Management request flows, routing users to the right request type based on their intent and collecting required application, environment, and role details in the same experience. For IT and access requests, Glean first attempts to resolve issues with knowledge, and when a ticket is still needed it creates a JSM request pre-filled with the correct fields so requests are triaged and routed correctly. This update is designed for organizations with complex, asset-backed JSM forms so agents and field teams can rely on Glean to submit tickets that match their real service desk process.
How to access: Administrators must enable for your organization. To use this capability, connect Jira Service Management to Glean and enable the Jira tools in the admin console under Platform and then Actions. Configure the Create Jira Service Management request tool with your service desk, request types, and required application, environment, and role fields. Update existing IT or access agents to call this action so that user requests are routed through the configured JSM workflow.
Set up Jira Service Management request routing
Actions [ROAD-1492]
GitHub sync and import for Skills
Admins can now import and sync Skills from public and private GitHub repositories into Glean. This makes it easier to manage shared Skill libraries in Git and keep Glean up to date with community, open-source, and internal Skills that follow the open Agent Skills standard.
Users can paste a GitHub repo URL, and Glean will fetch compatible .skill and .md files for import. Imported repositories stay up to date through a daily refresh, and users can refresh a GitHub-imported Skill before the next automatic refresh from the Skill's three-dot menu.
How to access: Administrators must enable for your organization.
Explore Skills in Glean Assistant
Assistant [ROAD-1407]
Skills for Glean Assistant: personal (public beta), governance and sharing
Glean Skills lets teams package domain-specific expertise into reusable instructions that Assistant applies automatically to tasks like account planning, meeting prep, and knowledge base creation. Skills support the open Agent Skills standard, so you can import compatible Skills from any public or private repository that follows the standard and tailor them to how your organization works. Admins can manage Skills rollout, sharing, and organization-wide governance with role-based controls, while personal Skills remain private unless users choose to share them.
How to access: Administrators must enable for your organization. Admins can manage shared and organization-wide Skills from the Admin console. End users can create personal Skills and use shared or admin-published Skills directly in Glean Assistant. Private Skills remain invisible to admins unless users choose to share them.
Explore Skills in Glean Assistant
Assistant [ROAD-1149]
Autodesk Construction Cloud (ACC) connector
Glean now includes a native Autodesk Construction Cloud connector, so users can search and discover ACC project files, RFIs, submittals, drawings, issues, assets, and related project content directly in Glean while respecting source permissions. This helps construction, architecture, engineering, and owner teams find project knowledge faster, discover related work across projects, and support compliance, audit, and onboarding workflows from one place. This is an admin-gated connector for end users.
How to access: Administrators must enable for your organization. In the Glean admin console, go to Data sources and add the Autodesk Construction Cloud connector. Follow the setup flow in the connector configuration page to authenticate Autodesk Construction Cloud and complete the required admin configuration. After setup is complete, run the initial crawl so ACC content becomes available to users with the appropriate source permissions.
Set up the Autodesk Construction Cloud connector
Connectors [ROAD-1326]
Panopto connector
The Panopto connector enables organizations to seamlessly integrate Panopto video content into Glean, making it easy to search, discover, and reuse knowledge stored in training videos, recorded meetings, and other video assets.
How to access: Administrators must enable for your organization. A Glean admin must add Panopto as a connector in the Glean admin console.
Set up the Panopto connector
Connectors [ROAD-1002]
Brave web search: freshness support for time-sensitive queries
Brave web search now scopes time-sensitive queries to fresher results, so questions about what is happening today, this week, or this month return up-to-date web pages instead of stale content. For customers using the Brave web search provider, Assistant only applies a recency window when the query clearly asks for current or recent information, improving relevance while keeping timeless facts unchanged. There are no new toggles or admin controls for this behavior; once Brave web search is enabled, freshness is on by default.
Configure Brave web search in Glean
Actions [ROAD-1514]
Claude Sonnet 5 in Assistant and agents
Claude Sonnet 5 is now available in Glean Assistant and agents.
How to access: Glean admins can enable Claude Sonnet 5 in the model hub.
This feature may be subject to usage-based pricing. Learn more in Glean's Help Center.
Agents [ROAD-1546]
Autonomous agent compiler: persistent chat sessions across page refreshes
Compiler chat persistence in the autonomous agent builder keeps your conversation history when you refresh the page or reopen an agent, so you don't need to re-explain context or repeat earlier steps. When you return to an agent, Glean automatically restores your latest compiler chat for that agent, and a Reset chat control lets you start a fresh conversation while previous sessions stay available in history. This improvement is available to all agent builders.
Learn about compiler chat persistence
Agents [ROAD-1513]
Artifacts support in agents
Enables artifacts in agents so they can generate rich outputs instead of only plain chat responses. Supports all artifact types that are enabled in Assistant.
This feature may be subject to usage-based pricing. Learn more in Glean's Help Center.
Work with artifacts in Assistant
Agents [ROAD-1458]
Navigate admin features more efficiently with a new streamlined left navigation and user menu
Updates to the left navigation and user menu now make it easier for admins to find valuable insights. The admin console now lives in the user menu, Insights has its own navigation path, and the updated user section improves admin accessibility.
Explore admin navigation and chat history
Assistant [ROAD-1467]
Take actions through real-time voice — send emails, post messages, create events, and more
You can now take actions directly through real-time voice — send emails, post Slack messages, create calendar events, file Jira tickets, and more, all by speaking. Say what you want to do, confirm with your voice, and Glean handles the rest. Voice actions work with all the same tools and integrations you already use in chat, so there's nothing new to set up.
This feature may be subject to usage-based pricing. Learn more in Glean's Help Center.
Explore real-time voice in Glean
Assistant [ROAD-1370]
O365 federated search: container filter facets and GST facet parity
O365 federated search now supports container filters for SharePoint and OneDrive, so users can scope queries to specific sites, libraries, lists, or folders and get precise, bounded live results. This closes a parity gap with indexed search and keeps behavior fail-closed instead of broadening the search when a container cannot be resolved. The feature applies to tenants using O365 SharePoint and OneDrive connectors with federated search enabled and requires admins to enable the container filter setting on those connectors and ensure live mode is turned on.
How to access: In the admin console, open each O365 SharePoint and OneDrive connector and enable the container filter option for federated search. Ensure O365 live mode is enabled for those connectors. After saving these changes, container-scoped queries from the Glean Search Tool and agents will automatically use O365 container filters.
Enable O365 federated search container filters
Connectors [ROAD-1459]
Enable Glean-managed MCP servers by default for customers
Glean Remote MCP Server is becoming much easier for admins to roll out and manage. Glean is enabling the Glean Remote MCP Server for customers by default with OAuth. This makes it easier for users to access Glean in supported AI tools without requiring admins to enable the remote MCP server manually.
How to access: The Glean Remote MCP Server is now enabled by default with OAuth for all customers. No manual admin setup is required. Users can connect to the remote MCP server from any supported AI tool using their existing Glean credentials.
This feature may be subject to usage-based pricing. Learn more in Glean's Help Center.
Enable the Glean Remote MCP Server
Connectors [ROAD-1397]
Choose search as your default setting in Assistant
You can now choose what happens when you press Enter in the Assistant chat bar. By default, pressing Enter still starts a new chat and Option/Alt+Enter starts a search. Users now have the option to change their personal settings so Enter starts a search and Option/Alt+Enter starts a new chat.
Customize your Assistant chat settings
Assistant [ROAD-1479]
Projects: organize chats and content in one place
Projects help you organize chats and content in Glean so related work is easier to group, revisit, and share. Create a project to keep relevant chats and artifacts together in one place. Projects are private by default and can be shared when needed. Existing Collections automatically migrate to Projects, with existing permissions preserved.
How projects work
Platform [ROAD-909]
Admin Features
Bulk agent cleanup: admin-driven inactive-agent identification and bulk delete (and Lifecycle API)
Admins can now clean up inactive agents at scale with a self-serve bulk deletion flow. From the admin console, go to Agents → Manage, where they can export an inactive-agents CSV, review which agents are unused, and soft-delete them in bulk from the admin console while preserving a 60-day restore window.
Delete and restore agents as an admin
Agents [ROAD-1453]

Original source