Results Query

What This Command Does

mcplab results provides machine-first access to evaluation run artifacts. Use it to find failures quickly with compact structured output, then fetch only focused context.

This is the recommended workflow for LLM and automation analysis over large result sets.

Recommended Workflow

• List runs to identify target run IDs.
• Search broadly with compact JSON output.
• Fetch focused scenario context only for top hit(s).
• Avoid loading full trace.jsonl unless needed.

npx @inspectr/mcplab results search "tool failed timeout" --status failed --limit 10 --format json
npx @inspectr/mcplab results context --run <run-id> --scenario <scenario-id> --source trace --around 42 --format markdown

Subcommands

• mcplab results list — list available runs.
• mcplab results show --run <runId> — show run as json or summary markdown.
• mcplab results index [--rebuild] — manual index build/refresh.
• mcplab results search <query> — search indexed results/trace/summary.
• mcplab results context --run <runId> --scenario <id> — fetch focused excerpt.

Search Options and Defaults

Defaults are tuned for LLM use: compact and structured.

• --runs-dir: mcplab/results/evaluation-runs
• --status: all
• --source: results,trace,summary
• --limit: 10
• --format: json

npx @inspectr/mcplab results search "timeout" --status failed --agent claude-haiku --scenario search-tags --source trace --limit 5 --format json

Automatic Index Refresh

mcplab results search auto-refreshes local index when run files changed since last index build.

You can still prebuild or force rebuild for CI or batch workflows.

• Index file: mcplab/results/.index/results-search.jsonl
• Manifest file: mcplab/results/.index/manifest.json

npx @inspectr/mcplab results index --rebuild