modbender Metrics Dashboard
Track your agent's operational health. Record events, count things, measure durations, and generate reports.
Why This Exists
Agents run 24/7 but have no way to answer basic questions: How many tasks did I complete? What's my error rate? How long do API calls take? Which skills do I use most? Without metrics, you're flying blind.
Commands
Record a metric
python3 {baseDir}/scripts/metrics.py record --name api_calls --value 1 --tags '{"provider": "openrouter", "model": "gpt-4"}'
Record a duration
python3 {baseDir}/scripts/metrics.py timer --name task_duration --seconds 12.5 --tags '{"task": "scan_skill"}'
Increment a counter
python3 {baseDir}/scripts/metrics.py counter --name posts_published --increment 1
Record an error
python3 {baseDir}/scripts/metrics.py error --name moltbook_verify_fail --message "Challenge solver returned wrong answer"
View dashboard
python3 {baseDir}/scripts/metrics.py dashboard
View metrics for today
python3 {baseDir}/scripts/metrics.py view --period day
View specific metric history
python3 {baseDir}/scripts/metrics.py view --name api_calls --period week
Export metrics
python3 {baseDir}/scripts/metrics.py export --format json > metrics.json
python3 {baseDir}/scripts/metrics.py export --format csv > metrics.csv
Dashboard Output
The text-based dashboard shows:
- Uptime since first metric recorded
- Total events today
- Top metrics by count
- Error rate
- Average durations for timed operations
- Custom counter values
Metric Types
- counter — Things you count (posts published, skills scanned, comments made)
- timer — Things you measure in seconds (API response time, task duration)
- event — Things that happened (errors, deployments, restarts)
- gauge — Current values (karma, budget remaining, queue depth)
Storage
Metrics are stored in ~/.openclaw/metrics/ as daily JSON files. Lightweight, no database required.
Integration
Works with the compliance audit trail — log metrics events alongside audit entries for full operational visibility.