May 28

Agents

Generate or edit field content with AI

Filling in configuration fields in the Agent Component and MCP Server Trigger can be time-consuming and hard to get right, especially without a clear starting point.

With AI-assisted fields, you describe what you need in plain language and the platform generates the content for you. Fields that support this feature display a (sparkles) button in the upper-right corner. Click it, enter your description, and choose whether to apply or discard the result.

This feature is available in the following fields:

• Agent Component: User Message, System Message, and JSON Schema definition.

• MCP Server Trigger: Tool Name, Tool Description, and Trigger Description.

To learn more, read the Agent Component and MCP Server Trigger documentation.

Payload templates for MCP Server Trigger tools

The MCP Server Trigger now lets you test your tools directly from the trigger configuration, with pre-filled payloads added automatically to the Execution Panel.

There are two ways to test:

• Test Tools List: adds a payload that lists all tools available on the trigger. Click the Test Tools List button to add it to the Execution Panel.

• Test Tool: adds a payload pre-filled with the input parameters of a specific tool. Click the three-dot icon next to the tool and select Test Tool.

In both cases, click Play in the Execution Panel to run the test.

Learn how to test MCP Server Trigger tools

Platform Improvements

Pipeline Name filter for Monitor pages

The Overview, Executions, and Pipeline Logs pages now include a smart autocomplete search for pipelines.

You can search by exact name or use a partial match. Results appear in a dropdown, making it easier to locate the right pipeline in large environments.

To learn more, read the Overview, Executions, and Pipeline Logs documentation.

Consumption-Based Model: Improved credit status tracking

On the Consumption-Based Model page, the Credits history table now displays the Expired status for credit that has been fully consumed, even when the contract period is still active.

This gives you an immediate and accurate signal that a credit balance has been exhausted, so you can act quickly to purchase additional credits before impacting pipeline executions.

To learn more, read the Consumption-Based Model documentation.

Connectors & Triggers

Scheduler Trigger now supports Globals for Cron expressions

A new Cron Global type is now available for use with the Scheduler Trigger. In this Global, you can define different cron expressions for Test and Production environments, for example, every 5 minutes in Test and once a day in Production.

To use it, reference the Global in the Scheduler Trigger using Double Braces expressions. Variable resolution happens at deployment time: when you deploy the pipeline, the system validates and applies the referenced Global to that environment.

Configure Globals in the Scheduler Trigger

HubSpot CRM connector

A new HubSpot CRM connector is now available, covering core CRM objects: Contacts, Companies, Tickets, Engagements, Associations, Properties, and Pipelines.

This connector provides a dedicated integration point for CRM workflows and does not affect the existing HubSpot CMS and Sales connectors.

Learn more about the HubSpot CRM connector

PATCH and DELETE methods in the Dynamics 365 connector

You can now use the PATCH and DELETE methods in the Dynamics 365 connector, giving you more control over how you manage records directly from your pipelines. This improvement lets you update existing records partially and remove them entirely.

Learn more in the Dynamics 365 documentation

Bug Fixes

• Retry compounds: Fixed an issue where legacy pipelines with inconsistent track references (onProcess / onException) in Retry like compounds caused HTTP 500 errors during execution.

• Do While connector: Fixed an issue where the "Interrupt Loop On Error" toggle appeared as activated by default in the configuration form but was not included in the payload sent to the backend unless the user manually toggled it off and on.

• Pipeline metrics: Fixed an issue where historical execution data in the Connector Latency chart was not displayed for idle pipelines when applying long time-range filters (such as 1 hour, 6 hours or 7 days).

• Agent Component:

  • Fixed an issue where configuring the Agent component with the GPT-5 model caused execution failures because an unsupported temperature value was being sent to the provider.

  • Fixed an issue where custom parameters (such as reasoning_effort) persisted when switching models and reappeared after being deleted, causing execution failures with unsupported argument errors.