July 28

Now in Early Access: Digibee Digital Workers

Delegate design, build, and more to integration-specialized AI that own the work end-to-end. You stay in control and oversee digital workers as they ship real, tested pipelines. When you log in you'll see a new option to switch between Manual Mode or Digital Worker Mode so you can choose how you want to work at any given time.

Learn more

Agents

Agent logs inside the Pipeline Logs page

You can now inspect and act on agent execution logs directly from the Pipeline Logs tab in Monitor:

• Agents view: See full details of each agent execution, from configuration to final output, directly in the Pipeline Logs tab.

See how it works

General experience improvements in the Agent Component

This release brings several smaller improvements to the Agent Component, from annotations and editing to output configuration.

• Annotations: Annotations now appear directly in the Output section of an execution, alongside Tools, System Message, User Message, and Guardrails. Click Add Annotation below the output to record why it worked or failed.

• Editing experience: System Message, User Message, and Evaluations prompts now expand automatically as you type, support manual expand and collapse, and include a copy button for quick reuse.

• Output Format: The Response Format section now lives on the main page of the Agent Component configuration, renamed to Output Format. Choose between Text or JSON Schema, generate a schema automatically with AI Assisted, and rely on automatic syntax validation. A tag always shows the format currently selected.

• AI Assisted character limit: The character limit for AI Assisted descriptions in the JSON Schema Definition field now goes from 1,000 to 10,000 characters, so you can write more detailed instructions when generating a schema.

• Custom parameters: Anthropic models now support custom parameters, giving you more flexibility when configuring model behavior.

Learn more in the Agent Component documentation

Connectors & Triggers

Byte-for-byte pass-through now available on the HL7 connector

The byte-for-byte message pass-through introduced for the HL7 Message Transformer is now also available on the HL7 connector. Enabling the new pass-through parameter skips the normalizing parse step, so no characters, escape sequences, or trailing empty fields are altered, ensuring the message reaches its destination unchanged.

Learn more in the HL7 connector documentation

Full DN bind support now available on the LDAP connector

The LDAP connector now supports binding with a full DN. By leaving the Authentication DN parameter blank, the connector uses exactly the full bind DN entered in the Username field in the account configuration on the platform.

This makes it possible to authenticate against directories whose RDN isn't CN, such as those using UID=, sAMAccountName, or a UPN like user@domain. These configurations are also validated at deploy time, catching errors like an empty username or unparseable DN before execution.

Learn more in the LDAP connector documentation

Bug Fixes

• MCP Server Trigger: Fixed an issue where pipelines exposed as MCP tools could incorrectly return a "Tool not found" or "Permission denied" error even when execution completed successfully.

• MCP Server Trigger: Fixed an issue where executions in the Execution Panel could fail with a "Permission denied" error caused by unnecessary token validation.

• Agent Component Execution: Fixed an issue where execution logs could show incomplete data and dataset test case executions could appear merged into a single execution.