list_runs({"jobId":"JOB_ID","limit":10}), then get_run({"runId":"RUN_ID"}). Current MCP does not return an output field or trace messages. Open the browser run page to read the trace, or retrieve REST trace pages with an authorized client. Report the state/link through MCP and only summarize findings after actually reading them.
Run pages return items and nextCursor; send that cursor for the next page. Optional state and contextQuery filter runs. Context search is a case-insensitive substring, not a query language. Current list rows contain no context excerpt; get_run returns full context, rendered prompt, invocation.trigger_id, execution metadata and activity.
Check that the intended trigger fired, the repository/task appears in the rendered prompt, and the browser/REST trace supports the reported result. A queued submission is not a successful execution. Browser run pages are at https://app.factorize.sh/job-runs/RUN_ID. Agent output stays in Factorize and is not automatically copied into Linear.
For missing output or execution failures, call get_run_diagnostics and follow failed-run troubleshooting.