ClawMetry = open-source observability dashboard for OpenClaw AI agents. Single Python file, zero config, auto-detects everything.
- Don't split dashboard.py — the single-file architecture is a feature, not a bug
- Don't add heavy dependencies — Flask is the only required dependency
- Don't add a build step — frontend is embedded in Python, no npm/webpack/vite
- Keep auto-detection working — if you add a feature, it should work without config
- Performance matters — this runs on Raspberry Pis. Keep it light (<100MB RAM)
- Read-only principle — ClawMetry observes agents, it doesn't control them
- Add a new API endpoint: Add
@app.routein dashboard.py, follow existing patterns - Update the UI: Edit HTML template strings in dashboard.py (search for the route's
render_template_string) - Add a new dashboard view: Add route + template + navigation link
- Bump version: Update
__version__in dashboard.py andsetup.py
See PUBLISH.md in packages/clawmetry/
- Don't introduce React/Vue/Angular — vanilla JS only
- Don't require a database for core functionality (history.py is optional)
- Don't make external network calls from the dashboard
- Don't store user data outside the local machine
- Don't break backward compatibility with older OpenClaw versions