Preview trace-to-eval capture for a run

Dry-run of POST /eval/suites/{id}/cases/from-execution: reconstructs the agent execution exactly as capture would (same reader, same message math) and returns its recorded actions with the fork index each one maps to, plus whether the case would be fully replayable. Reads only — nothing is written. Restricted to agent executions the caller owns.

Authentication

AuthorizationBearer

API key or Clerk session token

Path parameters

executionIdstringRequired>=1 character

Response

Capture preview

actionslist of objects

The recorded tool-call actions in invocation order — the fork-picker rows.

defaultForkMessageIndexinteger

The fork the server uses when forkMessageIndex is omitted: just before the last recorded action, or the end of the seed for a run with no tool calls.

executionIdstring

messageCountinteger

Length of the full reconstructed history (seed messages + one call/result pair per action).

replayableboolean

True when every recorded tool output would be captured in full (nothing truncated).

Errors

400

Bad Request Error

401

Unauthorized Error

403

Forbidden Error

404

Not Found Error

500

Internal Server Error