Skip to main content
Expose Autonomous Agent Graphs as callable endpoints. External systems or the platform scheduler invoke a pinned Agent Build per DEV / UAT / PROD.
Conversational graphs use Channels and Chat API — not these Autonomous modes. See Trigger types.

​
Execution modes

​
Enable a trigger

  1. Open Integrations → Triggers (or Deploy → Deploy as API / Cron job in Graph Studio).
  2. Select workflow type — API, Background Task, or Cron Job.
  3. Choose the Agent Graph and assign an Agent Build per environment.
  4. Save — the platform generates URLs for each environment.
Map builds per environment on Mapping triggers.

​
Authentication

All HTTP trigger requests require a workspace Bearer Token:
Test in DEV before promoting build assignments to PROD. Sample requests: API usage examples.

​
API mode (single endpoint)

Synchronous execution — the call waits for the graph to finish and returns the result.
Response status is completed or failed with workflow_id, response, and logs.
Not recommended for complex graphs. Use Background task for longer work.
Asynchronous start + status endpoints.

​
Start

Returns workflow_id with status: "pending".

​
Status

Poll until completed or failed. Implement retries with backoff on the caller side.

​
Cron job mode

Runs on a schedule inside Phinite — no external HTTP caller after configuration.
  1. Save the graph and create an Agent Build.
  2. Integrations → Triggers → Cron (or Studio Deploy → Cron job).
  3. Set the cron expression and default message / user_variables.
  4. Assign builds per environment; confirm runs in Observability logs.
Cron job schedule UI

Cron trigger configuration

Design Cron graphs to be idempotent — duplicate ticks should not double-apply side effects.