Amplitude Release Notes

Custom Agents are now live and generally available across all customers.

WHAT IT IS

Description: Global Agent is great for ad hoc questions, but what about the ones you ask every week? Custom Agents lets you create purpose-built AI agents tailored to your specific workflows, data, and questions — then let them run autonomously on a schedule, so the work happens without you having to ask. Configure each agent with its own instructions, scope, and tools. Simply connect Amplitude data alongside external apps like Jira, Linear, Slack, Granola, and Sentry to automate recurring tasks like backlog grooming, funnel analysis, churn detection, release monitoring, and more.

With this release:

• Build a custom agent in seconds using the /create-custom-agent skill in Global Agent
• Choose new agent templates like churn monitoring, feature adoption tracker, and more
• Connect external apps via Connectors to bring cross-tool context into every agent run
• Access all chat, specialized, and custom agents in one place under the new left-nav item

Availability: Access via Agents in the left nav → Create Agent.
When: Generally available now on all plans.

Description

Debugging an Event Streaming integration just got a lot easier. Live Log is a new real-time stream on the Event Streaming sync detail page that shows exactly what's happening as events are processed — so you can confirm the right events are flowing, how they're being transformed, and what the destination sends back.

For each event, inspect all three stages side by side:

• Input: the raw Amplitude event
• Dispatched: the transformed payload sent to the destination
• Response: the destination's status and response body

Pause and resume the stream, search across payloads, or open any row for full detail.

Where

Access via Data → Destinations → select an Event Streaming sync → Live Log tab.

When

Generally available now on all plans.

What: Agents can now trigger cohort syncs directly from MCP 6 closing the gap between observing cohort data and acting on it. Sync to Braze, Iterable, S3, webhooks, and more with a single tool call that previews and confirms before firing the export.

Combined with the existing get_cohort_sync_destinations, get_cohort_syncs, and get_cohort_sync_history tools, agents can now manage the full cohort sync workflow end-to-end without leaving your AI assistant.

Where: Access via the Amplitude MCP server or public MCP server.

When: Available now on all plans.

Description

AI assistants using the Amplitude MCP server now have a much richer view of your Guides and Surveys configuration. Previously limited to basic retrieval, the external MCP server can now fully understand a guide's published state, prioritization, and all other Guides and Surveys concepts — so AI-assisted workflows can reason about and act on your guides more accurately.

Where

🔓 Access via the Amplitude MCP server in ChatGPT, Claude, or any other MCP-compatible AI surface.

When

Generally available now on all plans.

Description

React Native Session Replay now correctly respects your chosen maskLevel — light, medium, or conservative — on both iOS and Android, honoring privacy configurations end-to-end exactly as intended. Bonus points? iOS light masking now also auto-masks phone and email keyboard inputs to match Android and web behavior.

With this update:

• Your configured maskLevel is respected end-to-end on iOS and Android
• Conservative masking on iOS masks React Native text views by default
• Light masking on iOS auto-masks phone and email keyboard inputs (now at parity with Android and web)

Requires:

@amplitude/[email protected] (standalone) or @amplitude/[email protected] (plugin) + AmplitudeSessionReplay iOS >= 0.11.1.

When:

Generally available now.

Description

You can now build, iterate on, and ship tracking plan changes entirely through AI-assisted workflows. The Amplitude MCP server now includes create and branch management tools for taxonomy, bringing a workflow that was previously only available in the Amplitude UI to any AI assistant.

With this update, you can:

• Create events, custom events, and properties via AI
• Propose and review tracking plan changes on a branch before merging
• Manage the full branch lifecycle — create, update, merge, refresh, and delete — without leaving your AI assistant

Where

Access via the Amplitude MCP server or public MCP server.

When

Generally available now on all plans.

Description

Ready? Set? Go! Global Agent now connects to the tools where your team plans, ships, and communicates — so you can ask cross-source questions and take action without leaving Amplitude. Simply connect to Slack, Atlassian, GitHub, Sentry, and more to ground analysis in real context alongside your Amplitude data.

With Agent Connectors, you can:

• Correlate analytics with product work | Example: Ask to compare a retention drop against Jira incidents or Linear bugs from the same sprint.
• Ground analysis in product specs | Example: Pull a Confluence spec into the conversation, then ask to find the matching funnel or cohort.
• Close the loop on insights | Example: Turn analysis into a Jira ticket, Linear issue, or Slack post without leaving Global Agent.
• Summarize cross-tool context | Example: Ask to combine a GitHub PR description, Linear ticket, and Amplitude chart into a summary.

Where: Access via Global Agent → Select Mode → Connectors → Add Connectors.

Available connectors at launch include:

• Atlassian (Jira + Confluence)
• Linear
• Notion
• Slack
• GitHub
• Granola
• Sentry
• Datadog (coming soon)
• Figma (coming soon)
• Plus support for custom connectors so you can bring your own MCP server.

Docs: https://amplitude.com/docs/amplitude-ai/global-agent-connectors

Description

As AI agents become a bigger part of how teams work with Amplitude, orgs need finer-grained control over what those agents can access and do. MCP sessions now respect role-based permissions at the project level — so you can scope exactly what an agent can read or write, without changing anything for users who don't need restrictions.

With this update, you can:

• Assign new READ and WRITE MCP actions to any role, user, group, or service account
• Restrict agent access on a project-by-project basis — agents can't touch projects they shouldn't
• Apply team-wide MCP access policies via group support
• Extend the same controls to service accounts for automated agent workflows
• Existing roles default to read and write enabled, so nothing changes until an admin explicitly configures restrictions.

Where

Access via Settings → Role Management → create or edit a role → toggle USE_MCP_READ / USE_MCP_WRITE in the AI/MCP section.

When

Available now on all plans.

Session Replay Agent v2 improves replay analysis by using video analysis to generate structured insights and combining it with autocapture and analytics for clearer, evidence-backed insights.

WHAT IT IS

Description: The latest version of our Session Replay Agent brings a fundamentally new approach to replay analysis. V2 watches actual session footage using multimodal video analysis, turns it into structured findings, then validates and quantifies those findings with autocapture and analytics data — producing clearer, more evidence-backed insights than ever before.

With this update:

• Video analysis gives the agent a deeper understanding of what users actually experienced
• Autocapture and analytics data validate replay findings, moving from qualitative reads to defensible insight
• Native mobile session replays are now analyzable — a major unlock for mobile teams
• The agent returns structured insights, highlight reels, quantified impact, and actionable recommendations tied to actual UI elements and flows
• The agent supports both interactive exploration and scheduled runs for proactive funnel and event monitoring

Availability: Access via AI Agents → Create Agent → Session Replay Agent.
When: Generally available now on all plans.

Description

Your CS team hears the real customer pain on every call — now that context shows up where PMs actually look. Pipe Granola meeting notes into AI Feedback so sales and success calls appear alongside your in-product feedback signal. Every note becomes a searchable, tagged insight so you can see what customers said, not just what they clicked.

Where

Access via AI Feedback → Sources → Add Source → Granola. Requires a Granola Enterprise or personal API key.

When

Generally available now on all plans.

Description

E-commerce hubs give merchandising and growth teams a curated starting point for analyzing product performance and product discovery in Amplitude. Instead of starting from a blank chart, teams get preconfigured commerce workflows with purchase funnels, conversion metrics, revenue, AOV, cart drop-off, product breakdowns, and discovery attribution built in.

Why

E-commerce teams constantly need to answer questions like:

• Which products and categories are driving revenue?
• Where are users dropping off in the purchase funnel?
• Which discovery paths actually lead to conversion?

Today, answering those questions often requires manually configuring multiple charts, stitching together funnels and tables, or rebuilding the same workflows repeatedly. E-commerce hubs streamline that process by combining curated workflows with the flexibility of Amplitude’s exploratory analysis experience.

Teams can start from high-signal commerce views and seamlessly drill deeper into tables, segmentation, cohorts, and activation workflows:

• The product discovery hub is powered by persisted properties, allowing teams to attribute downstream purchases back to acquisition and discovery methods like campaigns, search terms, or entry pages
• Purchase analysis is powered by cart properties and object arrays, enabling accurate attribution of conversion and revenue across products, brands, and categories.

Where

Available from the "Ecommerce Analytics" section.

Available in EUDC

yes

Plans

Available for Growth and Enterprise customers (with access to Cart Analysis / cart properties functionality)

What

The User Profile Timeline now surfaces frustration signals directly alongside behavioral events – error events with distinct styling, key properties inline, error-focused filtering, and quick links to Session Replay. This gives you a chronological view of exactly what a specific user experienced.

Why

Engineers and support teams investigating bug reports and user complaints now have a native home in Amplitude too. No more cross-referencing tools (and teams) to reconstruct what went wrong. It's all in one place on the User Profile.

Where

User Profile → Activity

When

Generally available now

What is Zoning Insights?

Every element on a site is either converting visitors or losing them. Figuring out which one used to take hours. Zoning Insights gets you there in minutes.

Features

• Select any element on any page to see real-time performance data, including clicks, exposure, and revenue impact
• Switch between overlays to find what's working and what isn't at a glance
• Filter by Amplitude cohorts to see how different user segments behave on the same page
• Drop high-intent users into dynamic cohorts for retargeting or experiments without leaving the workflow
• Use automatically mapped page groups to aggregate performance across entire templates like PDPs or blogs
• Get tailored recommendations you can test and target, right inside Amplitude
• Share zone reports with your team or jump directly from a zone into a journey or replay

Plans

All plans have access to Zoning Insights, and an add-on is available to access unlimited zones and overlays, revenue metrics, the Chrome extension, and more.

Description

Persisted Properties let teams preserve important behavioral context across a user journey so properties like campaign, entry page, search term, or discovery method stay attached to downstream events like purchases, conversions, and revenue. Instead of analyzing isolated events, teams can carry attribution context across the full customer journey.

Why

Teams constantly need to answer questions like:

• Which campaigns actually drive conversion?
• Which landing experiences influence downstream revenue?
• Which discovery paths lead to purchases?
• Which acquisition channels bring high-intent users?

Today, answering these questions often requires rebuilding attribution logic manually in charts, SQL pipelines, warehouses, or duplicated instrumentation. Persisted Properties make this native inside Amplitude by allowing teams to configure attribution behavior directly in the product and reuse it consistently across analyses.

Where

Available in Data → Properties → Persisted Properties. Persisted Properties can then be used directly across charts, funnels, tables, cohorts (and e-commerce hubs!).

Available in EUDC

Yes

Plans

Available to all.