Skip to main content

Agent Executions API

Inspect and stop individual agent executions.

📄️Get execution

Returns the execution detail. Executions that already finished or stopped are served from a cache, running ones are always read fresh, so this endpoint can be polled to track a live execution. The `applicationId` query parameter is optional: when omitted, it is derived from the execution itself. Node detail is progressive: `fields=outline` answers 'which nodes ran, in what order' in a few hundred bytes, `fields=minimal` (the default) returns a lean, de-duplicated timeline, `fields=compact` keeps full node output but drops the node’s heavy `input` snapshot, and `fields=full` returns everything including each node’s `input.context` (with environment values masked, and a forEach body node carrying its wave’s scope rather than its own item binding — the items are on the forEach node’s `output.items`); `nodeIds` narrows to specific steps. Document URLs inside node outputs and the input message carry a time-limited `vid` token, so they can be fetched, rendered or attached without platform credentials for a few hours (re-fetch the execution for fresh ones). Use `POST /api/v1/agents/{agentId}/executions/{executionId}/stop/` to cancel a running execution.