gatus-mcp
An MCP server that exposes a
Gatus status page to AI agents. It wraps Gatus's
REST API and serves tools over streamable HTTP at /mcp, so an assistant can
answer "what's up / what's down" and drill into uptime, response times, and probe
history.
Monitors are defined in Gatus's own YAML config — this server does not create
or change them. Every tool is read-only except submit_external_result, which
pushes results for endpoints Gatus is configured to treat as external.
Tools
Windowed tools accept a duration of 1h, 24h, 7d, or 30d.
The server also exposes GET /health ({"status":"UP"}) for docker/k8s liveness probes.
Configuration
All configuration is via environment variables:
Running
Docker
docker build -t gatus-mcp .
docker run --rm -p 3000:3000 -e GATUS_URL=http://your-gatus:8080 gatus-mcp
Docker Compose
To run Gatus and gatus-mcp together, copy
docker-compose.example.yml to
docker-compose.yml, supply your own gatus.yaml, and run docker compose up.
From source
go run ./cmd/gatus-mcp
# or
go build -o gatus-mcp ./cmd/gatus-mcp && GATUS_URL=http://localhost:8080 ./gatus-mcp
The MCP endpoint is then available at http://localhost:3000/mcp.
Connecting a client
Point any streamable-HTTP MCP client at http://<host>:3000/mcp. For example,
with Claude Code:
claude mcp add --transport http gatus http://localhost:3000/mcp
Development
go test ./... # unit + parity tests
go vet ./...
golangci-lint run # config in .golangci.yml
Layout
cmd/gatus-mcp/ entrypoint: env config + HTTP serving
internal/gatus/ REST client for the Gatus API
internal/mcpserver/ MCP tool definitions and handlers
License
MIT © Adam Benhassen