Skip to main content
Feedback

AI Release Notes for May 2026

Platform Release

May 30 2026

Agentstudio

New
We added these features

  • The MCP Registry is a centralized hub in API Control Plane for governing which MCP servers your organization can use. It aggregates servers from three sources, the official Anthropic MCP registry and the internal Boomi registry. Boomi API products marked as MCP enabled automatically appear in the Boomi registry, requiring no manual addition.

    The Server Catalog in API Control Plane acts as the bridge between the MCP registry and Agentstudio. When a platform admin adds an MCP server to the Server catalog in the API Control Plane, it becomes immediately available inside Agentstudio's MCP Source configuration.

    To add an MCP Source from the registry, go to the Agent Designer's Source tab and click Add MCP Source. From there, you can browse or search the catalog and select a server. Credentials are to be manually configured by you. Connection details such as MCP server URL and any available headers are pre-populated automatically. This means your custom-built agents get access to the right capabilities and operations faster.

    API Products enabled as MCP server in API Control Plane require an AI agent key from ACP to authenticate. External MCP servers (such as those from the Anthropic registry) may not require additional authentication depending on how they're configured. Refer to Using the MCP server catalog. (BI-12837)

  • You can now access chat API Control Plane endpoints in Boomi GPT without encountering CORS errors. (BI-13246)

Fix
We fixed these issues

  • When you view endpoints from API Control Plane in Sources and change the view from 20 to 50 items, Agentstudio correctly shows 50 endpoints. (BI-12664)

  • Accelerator agents in Agent Garden now display a “Contact Us” option instead of “Install”, preventing failed installation attempts and aligning with the Marketplace experience. (BI-12674)

  • When you start a new agent conversation, responses now appear correctly in the UI. Previously, in environments with network proxies, a regression introduced in the May 2 release caused new conversations to appear stuck with no response surfacing. (BI-13531)

  • Agent streaming responses now include standard SSE headers, so intermediate progress notifications such as "thinking" and "calling tool" display as expected in environments with network proxies. (BI-13579)

  • Broken "Learn More" links on Agent Garden multi-region pages are fixed and point to the correct URLs, allowing you to access relevant help documentation. (BI-13472, BI-13640)

  • System guardrails no longer incorrectly block harmless prompts, reducing false positive policy violations during agent interactions. (BI-12258)

  • Tool calls that exceed the output token limit now return a clear user-facing message instead of exposing parsing errors or incomplete function call responses. (BI-13219)

  • Agentstudio now correctly generates PDF and YAML exports for agents that use MCP tools from MCP servers with titled endpoints. (BI-13649)

  • When a user message containing base64-encoded content triggers a content filter on the primary model, the request now automatically routes to a fallback model, returning a valid response instead of a silent failure. (BI-13500)

  • AWS Bedrock Guardrails for agents now apply prompt attack filtering only, reducing false positive guardrail triggers and improving user experience while maintaining safety with prompt attack filter and also keeping all local guardrail configurations (word filter, regex, denied topics) active. (BI-13501)

  • Standardizes Trust Level terminology across the Agent Control Tower UI, ensuring consistent labeling as "Endorsed," "Unendorsed," or "Deprecated" for a cohesive user experience. (AGCT-5336)

  • Anomaly Summary cards now display category-specific icons and a corrected Peak Anomaly Day date format, making anomaly details easier to scan and interpret. (AGCT-4776)

  • Anomaly table headers and time range formats are now standardized, improving consistency across tables and overall usability. (AGCT-4394)

  • The information icon on the Anomaly page is now correctly positioned, with duplicate definition text and tooltips removed for a cleaner interface. (AGCT-4393)

DataDetective

New
We added this feature
  • The DataDetective agent now supports pagination with sorting and filtering, allowing you to view each process and data field while seamlessly navigating large datasets. Refer to Boomi DataDetective to learn more. (BI-13183)

Boomi GPT

Fix
We fixed this issue
  • Requests to “Create a DataHub model” in Boomi GPT now correctly route to the HubGen agent. (BI-13638)

May 2 2026


Agentstudio

New
We added these features

  • Agentstudio multi-region instances allow you to deploy agents across regional runtime clouds based on your data residency and compliance requirements. Agent execution stays within the selected region, helping you enforce AI boundaries and reduce latency.

    You can package and deploy agents to environment-specific runtimes across USA East, UK, Japan, and Australia Agent Garden Clouds. You can also deploy the same agent across multiple regions with full visibility into what is running and where.

    Agent Control Tower provides a unified view to track and monitor all agent deployments across runtimes, simplifying governance and oversight.

    Your existing agents continue running without interruption, defaulting to USA East Agent Garden Cloud, or GBR Agent Garden Cloud if you're on the European Boomi platform. Refer to Deploying agents to learn more. (BI-12889, AGCT-3707, AGCT-4548)

  • You can now ingest Claude Code's OTLP telemetry (token usage, cost in USD, active time) into Agent Control Tower allowing you to track metrics on the monitoring page. Refer to Enabling metrics for Claude Code to learn more. (AGCT-5089)

On this Page