Observability
AgentOS exposes free local observability by default. No hosted monitoring service is required.
Prometheus Metrics
Metrics are served from the admin port:
Local Prometheus scrape config:
The main gateway metrics are:
| Metric | Meaning |
|---|---|
AgentOS_requests_total |
Request count by tenant, method, path, and status |
AgentOS_request_duration_seconds |
Request latency histogram by tenant, method, and path |
Grafana
Dashboard JSON:
Import the dashboard into a local Grafana instance and point it at a local Prometheus datasource. The dashboard includes:
- gateway throughput
- p50 and p95 latency
- request count by status
- request count by tenant
- request count by path
Local-Only Docker Example
Run AgentOS:
Run Prometheus locally with the included config:
docker run --rm -p 9090:9090 \
-v "$PWD/deployments/observability/prometheus.yml:/etc/prometheus/prometheus.yml:ro" \
prom/prometheus:latest
Grafana can be run locally as well:
These containers are optional. AgentOS itself exposes metrics without them.