See how a coding agent made its pick.
Replay a finished Agent Picks run to see what the coding agent searched, fetched, ran and wrote, the tool it picked and what backs the pick.
Run console
Read a finished run’s transcript in order: what the agent searched, read, fetched, ran and wrote. Fetches show their HTTP status, so a 404 on your docs stands out.
Why this pick
See the tool a run picked and what backs it: the packages the agent added, the package manager it used and the files it changed.
Stated picks and installs
Some runs pick a tool in their final answer without installing it. The pick rate counts these runs; the install rate beside it counts only runs that installed the tool.
Run history
Every run is listed with its test, agent, model and status. Open a finished run, watch a running one live, or see why one failed.
API and MCP
List runs and read a run’s trace over the REST API, or have your own agents inspect a run’s pick and recorded activity over the MCP server.