What you get
- 🚨 One queue — alerts from every source, deduplicated by ID, with owner, status and comments on each
- 🔎 Filter, search, group — by severity, source, tag, team, time window; save the view as a dashboard
- 🔕 Mute policies — scheduled or ad-hoc silences that match on any alert field, without deleting anything
- 🏷️ Enrichment rules — add tags, links and runbooks to alerts as they arrive, by pattern
- 🧠 Root cause analysis — your system pushes a root cause per alert via the API; operators rate it 👍/👎 and the verdict is relayed back to the sender
- 📈 Insights — MTTR, MTTA, volume and re-fire trends, per tag, with hourly resolution on short windows
- ☎️ On-call teams and role-based access (admin / editor / viewer)
Key Features
🚨 Alert sources
Every integration is a webhook — point your tool at OpsiMate and alerts appear in the queue:
Each source has a setup guide in the app (Integrations → Add) with the exact webhook URL to paste.
Run it
Docker — one command, nothing to clone
curl -fsSL https://raw.githubusercontent.com/OpsiMate/OpsiMate/main/scripts/start-docker.sh | sh
Access the application:
- Backend: http://localhost:3001
- Client: http://localhost:8080
- API reference: http://localhost:3001/api/docs — Swagger UI with "Try it out"; the OpenAPI document is at
/api/openapi.json (set API_DOCS_ENABLED=false to turn both off)
- AI assistants (MCP):
http://localhost:3100/mcp — the mcp container, a read-only MCP server for Claude Code, Claude Desktop, Cursor and other assistants: dashboards, alerts, comments, root causes and analytics through seven tools, read from the backend with each caller's own token. Connect with claude mcp add --transport http opsimate http://localhost:3100/mcp --header "x-api-token: <token>"; Settings → AI has snippets for other clients. Don't want it? Remove the mcp service Existing installs: fetch the latest docker-compose.yml first — scripts/start-docker.sh keeps the one you have, and older files have no mcp service.
The first account you register becomes the admin.
Kubernetes
A Helm chart lives in infrastructure/helm; Terraform for the surrounding infra is in infrastructure/terraform. See the deployment docs.
Volume Mounts (optional but recommended)
| Volume | Purpose | Required For |
|---|
/app/data/database | SQLite database persistence | Backend + Worker |
/app/data/private-keys | SSH private keys for authentication | Backend + Worker |
/app/config/config.yml | Custom configuration | Backend + Worker |
Configuration
server:
port: 3001
host: "0.0.0.0"
database:
path: "/app/data/database/opsimate.db"
security:
private_keys_path: "/app/data/private-keys"
# Machine-to-machine token: your monitoring tools send it as X-API-Token
# when pushing alerts and root causes. Change it.
api_token: "change-me"
Set ENCRYPTION_KEY (credentials at rest — the server warns at startup if it's missing) and JWT_SECRET (session signing) in the environment for anything beyond a laptop.
Contributing
New here? Start with a good first issue — each one names the exact file and line, and a maintainer reviews within a day or two.
pnpm install
pnpm dev # client on :8080, server on :3001
pnpm run check # lint + prettier — run this before you push; CI runs the same
pnpm test
Bigger areas where help is welcome: new alert sources, the incident-management work, and the AI investigation agent (see roadmap).
Roadmap
Recently shipped
- 📈 Alert Insights — MTTR / MTTA / volume trends, hourly on short windows
- 🧠 Root cause analysis via API, with operator ratings fed back to the sender
- 🤖 AI-powered alert search
- ⚡ Server-side query engine — the queue stays fast at 10K+ alerts
Next
- 🤖 Built-in investigation agent — one-click root cause that actually looks at your cluster, logs and metrics (HolmesGPT integration, in design)
- 🔄 Incidents — group related alerts into a named incident with a timeline and postmortem
- 📣 Alert routing — escalation to Slack / on-call by severity and team
Support
💖 Our Amazing Contributors
This project wouldn’t be what it is today without the incredible people who have shared their time, knowledge, and creativity.
A huge thank you to everyone who has helped and continues to help make OpsiMate better every day! 🙌
