Boomi AI Updates & Release Notes

Platform Release

Agentstudio

New Features

• Introduced automatic registration of the default Boomi Garden account in Agent Control Tower, eliminating the need to manually add an account, enter connection details, or trigger a sync. Agent data is now updated in real time in Agent Control Tower as changes occur in Garden, replacing the previous 30-minute sync. For more information, refer to Boomi Agent Garden Provider. (AGCT-6487, AGCT-5429)
• MCP source headers now include an Is Secret toggle, so you can mark a header as sensitive instead of relying on backend detection. Boomi masks the value of any header marked sensitive in the UI. Refer to Connecting to an MCP server to learn more. (BI-15540)
• The API tool now includes an Is Secret toggle for each header, so you can mark headers as sensitive instead of relying on backend detection. Boomi masks sensitive header values in the UI and retains your selection when you save, reopen, or run requests. Refer to Creating an API tool to learn more. (BI-15257)
• Introduced a full read-only view of agent design details for deployed agents, including model configuration, tools, tasks, guardrails, package, and deployment details. Open it by selecting View Details in the agent details panel. You can now also export packages as well as agents. Refer to Viewing agent details to learn more. (BI-14718,BI-15214, BI-15287)
• Extended the Agent Garden export API to support packages. The export endpoint returns a ZIP file with a YAML definition and a PDF, and export/preview lists the files before you download. Refer to Agent Garden REST APIs to learn more. (BI-15347)
• A single Agent Garden account can now be onboarded to multiple Agent Control Towers, removing the previous one-to-one constraint. (AGCT-8289)
• Introduced a new redirection link in Agent Control Tower connecting to dedicated agent pages in Agent Garden, allowing you to view complete details and configurations of an agent directly in Agent Garden. (AGCT-7187)
• Session logs are now available for Agent Step executions. You can view session logs for Garden agents deployed to the Boomi Cloud runtime, for both structured and conversational agents, regardless of whether they were invoked from Garden UI, the API, or Agent Step. (AGCT-6990, BI-14085)
• Admin users can now view session logs for all users in the account who have invoked the agent, complete with user information and source shown on each session, plus enhanced filtering. (AGCT-7256, BI-14092)

Enhancements

• You can now deploy agents to Boomi-managed clouds using your own model provider (Bring Your Own Model). Refer to Using your own models with agents to learn more. (BI-15005)
• Extended file upload support to agents that use your own models (Bring Your Own Model) in Boomi-managed clouds. Supported file types depend on the model you use. (BI-15441)
• Updated the Agent step to set Runtime Type automatically when you generate a configuration. Agents on a Flex Runtime use the Common setting and do not require a Platform API token. Refer to Configuring the Agent step to learn more. (RAS-989)
• Agent Garden now shows a message directing you to your account administrator when you try to add a Knowledge Base without a Knowledge Hub role, and there is no Knowledge Base available. Refer to Connecting a knowledge base to your agent for details on required roles. (BI-16215)

Fixes

• Resolved an issue where reopening a conversation from the conversation history showed the uploaded file attachment on every message in the session. The attachment now displays only on the message where you originally uploaded it. (BI-15659)
• Resolved an issue where MCP registry server names longer than 32 characters appeared truncated in the UI. Server names now display in full, matching what you configure in the registry. (BI-13914)
• Resolved an issue where deployed integration processes appeared with blank names in the Integration tool. Processes now display their name, or their process ID if Agent Garden cannot retrieve the name. (BI-15809, BI-15683)
• Resolved an issue where the Configuration tab in Agent Designer did not warn you about an incomplete model configuration after you deleted or changed its integration runtime environment. The Configuration screen warns about the change and Agent Designer disables the Create button until you reconfigure the model connection. (BI-15563)
• Resolved an issue in the UK region where AI-assisted features in Agent Garden, such as Build with AI and the Give me some ideas suggestions, failed. (BI-15640)
• Resolved an issue in Agent Control Tower where the agent status summary bar displayed counts for deploying and undeploying agents without labels. The summary bar now shows "Deploying" and "Undeploying" labels, consistent with other statuses like Active and Draft. (AGCT-7222)
• Resolved an issue where the agent list did not refresh after you imported an agent, so the imported agent did not appear until you reloaded the page. (BI-12676)
• Resolved two issues in Agent Garden: dates now display in your local time zone, and the tool row stays visible after you select Requires approval on the Tools tab. (BI-16059)

Agentstudio

Early Access

• Introduced Decision Engine mode, powered by Jev, for Agentstudio agents to make fast, schema-based decisions optimized for integration and automation workflows. Decision Engine agents accept structured JSON input and return validated, structured output, which eliminates the need for natural-language parsing in downstream systems.

Built for use cases such as invoice exception triage, access request gatekeeping, lead qualification, and claims triage, Decision Engine mode delivers faster performance and lower cost compared to conversational reasoning on the same tasks. It is currently available for agents deployed to Boomi-managed runtime clouds. Refer to Using Decision Engine mode for details. (BI-15769)

Agentstudio

New Features

• Providers like Langchain, CrewAI, and LangGraph to name a few now appear as their own provider tiles instead of being grouped under Custom Accounts, with accurate account and agent counts and correct name display on hover. You can add accounts directly from these tiles. For a list of all providers, refer to Other providers.
• The renamed Provider Accounts page (formerly Manage Providers) offers a clearer accounts list side drawer with a Close button, an info tooltip, and direct navigation to a provider's accounts list. That drawer now also lists every agent under a provider or sub-provider, and includes Custom API accounts alongside native entries for AWS Bedrock, Salesforce, Copilot, and Cortex, giving you one complete view of every account and agent tied to a provider.
• You can now add provider agents directly at the account level, not just within individual environments, either manually or by uploading through AI Catalog. When adding a Salesforce, Copilot and Snowflake Cortex, you can choose between Boomi Sync, which pulls data on a schedule, and API Push, which lets your systems push data to Boomi. For more information, refer to Choosing a connection method.
• Provider agent screens use Deployment instead of Alias across all labels, buttons, and headings for clearer terminology. This release also includes broader reliability improvements to custom provider elevation, resolving minor issues you could encounter when elevating custom providers. (AGCT-6963, AGCT-6863, AGCT-6817, AGCT-6743, AGCT-6710, AGCT-6709, AGCT-6708, AGCT-6683, AGCT-6565, AGCT-6564, AGCT-6360)

Platform Release

New Features

• Introduced a Create New Tool drop-down in the Available Tools drawer in Agent Designer, so you can build a missing tool without leaving your setup flow. Finish in a new tab, then click Reload to add it. Refer to Building an agent. (BI-14184)

Enhancements

• Added conversation starters to the Test Panel as clickable prompt pills, so you can explore agent capabilities without typing a prompt from scratch. If you already added conversation starters in the Profile section, the Test Panel pre-populates them automatically. (BI-14186)
• Agentstudio now supports connecting to a specific remote when an MCP server offers multiple remote options. For more information, refer to Using the MCP Server Catalog. (BI-14985)
• Administrators (having Role = Platform Administrators or Garden Administrator) can now view session logs for conversational agents from all users in Agent Control Tower. For more information, refer to What permissions do I need to view and interact with AI agents. (BI-14072)

Fixes

• Resolved an issue in the API Tool Builder where pasting text over 5000 characters into the Request Body field truncated the text and disabled the Save button. The Request Body field now truncates the text and lets you save. (BI-14256)
• Resolved an issue where the API tool configuration did not correctly map the authentication type from an uploaded JSON or YAML file. Uploaded specs now correctly populate the authentication type, including basic authentication, token, platform JWT, and OAuth 2.0/2.1. (BI-14716)

Agentstudio

New

We added these features

• Enhanced the Associations tab in Agent Designer’s glossary details panel to display agent title and mode for agents and model name for Data Hub models. Refer to Connecting a business glossary to your agent to learn more. (BI-14508)
• The Agent Garden Design API now offers documented /export and /import endpoints, where export returns an agent's YAML and PDF configuration as a ZIP file, and import accepts YAML only. For more information, refer to Boomi Agent Garden Design APIs - Agents. (BI-14490)

Fix

We fixed these issues

• Resolved an issue where the model configuration and file upload details were missing from the Agent Designer side panel when viewing package details. (BI-14383)
• Resolved an issue where the conversation history did not retain feedback selections made during a live session. (BI-14444)
• Resolved an issue where incorrect or missing error messages appeared when uploading files to an agent. (BI-14390)
• Resolved an issue where the Send button became disabled in the test chat interface after adding a file attachment and clearing the text input field. (BI-14384)

Agentstudio

New

We added these features

• Improves performance and responsiveness on the Agent Registry page in Agent Control Tower when displaying large numbers of agents, preventing crashes and allowing you to manage and view agent details at scale. Clusters agents by provider when more than 100 agents exist and provides a paginated list view per provider, while retaining the bubble view for 100 or fewer. You can zoom in on provider clusters to view individual agent bubbles and access their details. The same clustering, pagination, and zoom apply in the model view. (AGCT-5731)
• Groups model bubbles for OpenAI and Anthropic in Agent Control Tower, allowing you to visually distinguish and organize models by source and text similarity. (AGCT-6199)

Platform Release

New

We added these features

• You can attach Meta Hub business glossaries to your AI agents to give them access to enterprise terminology. Glossaries help agents better understand your organization's language, improving response consistency, accuracy, and alignment with enterprise standards.

  You can attach one or more published glossaries to agent tasks during agent design, giving agents business context such as definitions, acronyms, and relationships between key business entities. Refer to Connecting a business glossary to your agent to learn more. You can track your Meta Hub glossaries usage in Agent Control Tower through the monitoring screen, agent listing page and session logs. (BI-12940, AGCT-4543)

• Agent Designer now integrates with Knowledge Hub, enabling AI agents to securely retrieve and use enterprise knowledge from a unified, governed source.

  You can attach knowledge bases to agent tasks during agent design, allowing agents to ground responses and automations in trusted organizational data. This integration improves response accuracy, reduces hallucinations, and simplifies the creation of knowledge-aware agents. Using the Agent Control Tower, you can track Knowledge Hub tool invocations and retrievals through the agent listing page and monitoring screen. Refer to Connecting a knowledge base to your agent (BI-12843, AGCT-4545)

• The agent conversational interface streams an agent’s execution plan in real time during conversations when Extended Thinking is enabled. Instead of waiting on a static “Thinking...” message, you can follow the agent’s reasoning as it develops, improving visibility into agent behavior. Plan streaming displays in test and production modes. Refer to Using AI agents and Testing and troubleshooting agents to learn more. (BI-13986)

• Agents can now retain and recall information from previous tool responses throughout a conversation session, even when that data is no longer in the active context window. This session-scoped memory helps reduce redundant tool calls, improves reasoning over previously retrieved data, and enables longer, more data-rich conversations. Refer to Using AI agents and Agentstudio token limits to learn more. (BI-13790)

• Session logs now show Knowledge Hub steps in the Chain of Thoughts view including knowledge base and glossary retrieval steps, their counts, and any errors — and the Session Metrics panel shows Knowledge Base query and Meta Hub glossary call counts. Uploaded files appear as file-name chips in the Trace Input user prompt, and long user input and agent messages are truncated in the session invocation table tooltip. (AGCT-5842)

• The Add Source button is unavailable while creating a new API Control Plane source if the account does not have the required access. (BI-13799)

• Adds support for displaying disabled Salesforce agent versions in Activity Logs, maintaining a complete audit trail for multi-version scenarios. (AGCT-3494)

• The Add Source button is unavailable while creating a new API Control Plane source if the account does not have the required access. (BI-13799)

Fix

We fixed these issues

• OpenAPI tools configured with the PATCH method now execute successfully, allowing agents to invoke PATCH endpoints and receive API responses as expected. (BI-13937)

• Improves the Agent Listing experience (AGCT-5660)

Fix

We fixed these issues

• Boomi GPT correctly responds to prompts for DesignGen and Integration Advisor in Japanese and orchestrates the appropriate agent to perform the task. (BI-13797)

• This fix resolves image display issues in Boomi GPT responses for certain Boomi Answers intents, ensuring you can view relevant visual information. (BI-13605)

Fix

We fixed this issue

• This fix improves the Boomi Resolve functionality in the eu-west-2 region of EU Prod, enabling you to troubleshoot issues effectively. (BI-12810)

Agentstudio agents now support file uploads, enabling you to send documents and images directly to agents. This unlocks file-centric use cases such as invoice reconciliation, contract review, and data extraction from scanned images, documents, and spreadsheets.

Send files to an agent in the following ways:

• Drag and drop files or attach them in the Agent Garden conversational UI
• Send files through an agentic workflow in Integration with Agent step
• Include files in API requests

Supported formats include PDF, DOCX, XLSX, CSV, TXT, PNG, and JPG, with up to 5 files accepted per request. Individual files cannot exceed 2MB and total files cannot exceed 10MB. File upload support is controlled by an agent-level setting, which is off by default, so existing agents and text-only workflows are unaffected. Refer to Enabling file uploads to learn more.