Status pages backed by real metrics.
Observer is a metrics-driven status page platform. You define what "healthy" means as numeric thresholds on signals from your own infrastructure. Observer evaluates them continuously and turns each verdict into public status pages, SLOs, and incident timelines your customers can trust.
Point Observer at the signals you already trust. Each check evaluates against strict thresholds and produces one status.
- Metrics. Prometheus, OpenTelemetry (OTLP), CloudWatch, and direct active
probes: HTTP, TCP, DNS, TLS certificate expiry, gRPC, WebSocket, ICMP, host.
ICMP needs the standalone agent binary or a custom image, because the
published container image does not include
ping(ICMP probes). - Logs. Turn log volume and patterns into status with Loki and Elasticsearch aggregations.
- Databases. Probe Postgres, MySQL, Redis, and MongoDB directly with a read-only query.
- Agent. A self-hosted data plane that runs the checks inside your network and pushes verdicts only.
Turn those verdicts into something people act on.
- Status pages. Public or access-controlled, with 30-day history and uptime.
- SLOs. Targets, error budgets, and burn-rate timelines per service.
- Incidents. Post updates your customers actually read.
- Notifications. Reach subscribers where they are.
Pages can be scoped per customer for customer-facing status, or kept private for on-call and internal teams.
Two properties shape the design:
- Your telemetry stays yours. The agent
runs inside your network, evaluates each check locally, and pushes only the
result (
metric_id,value,status,timestamp). Raw query results, credentials, and connection strings never leave your network. - Your configuration lives in version control. The
CLI applies a single
observer.yamlthrough a workflow you own, so your setup is reviewed in pull requests instead of locked inside a UI.
your infrastructure open source (this org) use.observer (cloud)
┌──────────────────────────────┐ probe + ┌───────────────┐ push ┌─────────────────────┐
│ Prometheus, OTLP, CloudWatch │ evaluate │ Observer │ verdict │ Observer Cloud │
│ HTTP, TCP, DNS, TLS, gRPC, │ ─────────────▶│ Agent │ ──────────▶│ status pages, SLOs, │
│ databases, Loki, Elastic, … │ └───────────────┘ │ incidents, alerts │
└──────────────────────────────┘ └─────────────────────┘
┌───────────────┐ apply ▲
observer.yaml in your repo ─────────────────▶│ Observer CLI │ ─────────────────────┘
└───────────────┘
| Repository | What it is |
|---|---|
| agent | The data plane. Probes every source above, evaluates each threshold locally, and pushes verdicts over a single outbound HTTPS connection. Apache-2.0. |
| cli | Configuration as code. A dependency-free client over the Observer public API: plan changes on pull requests, apply on merge, export current state to bootstrap a repository. Ships a GitHub Action. ISC. |
| demo | A self-contained compose stack that exercises every probe source against real local services: one docker compose up, one observer apply, sixteen live metrics. The fastest way to see Observer working end to end. |
| skills | Agent skills and MCP setup: teach Claude, Cursor, or any MCP client to read your metrics and run your status page. |
- Create an account at use.observer.
- Install the agent next to your signals and connect it with the key from the Agents page.
- Define your first check, then publish a status page.
Want to see it running first? The demo stack brings up every probe source locally with one compose file.
Full walkthroughs live in the documentation. Prefer
config as code? Start with the CLI and
commit an observer.yaml.
- Product: use.observer
- Documentation: docs.use.observer
- Agent: github.com/useobserver/agent
- CLI: github.com/useobserver/cli
- Demo: github.com/useobserver/demo
- Skills: github.com/useobserver/skills
- Live status: status.use.observer
- Security: report vulnerabilities privately to security@use.observer