loom runs

Find saved workflow runs from the repository root or any subdirectory:

loom runs
loom runs --json

The table shows the exact run ID, workflow ID, last recorded workflow status, optional exit code, and first/last complete workflow timestamps with date and zone. - means unavailable. Rows sort by the first recorded workflow timestamp, newest first; ties use ascending run ID and unavailable starts come last. Run IDs and filesystem modification times do not determine recency.

A pending or running status is the last recorded state, not a claim that a process is still alive. Empty state files and runs with no complete workflow record still appear with their run ID and unavailable summary fields.

Choose an ID explicitly, then inspect its failed job or command:

loom logs <run_id> --state
loom logs <run_id> --job build --step build:script --command script-01

--json prints one summary per line using run_id, optional workflow_id, status, exit_code, first_workflow_timestamp and last_workflow_timestamp. Absent data is omitted; exit zero remains zero. If no runs exist, human output says No recorded runs. and JSON mode prints no lines.

Discovery reads only existing state.jsonl files. Incomplete final appends follow the normal reader rules; malformed complete records, missing or unreadable candidate state files, escaping paths and nonregular files fail the command before any rows are printed. Non-directory entries, symlink aliases and invalid run-ID directory names are ignored. Discovery writes no files and has no implicit latest selection, filters or limits.

See loom logs, Runtime logs and MCP loom_list_runs.