ops

Ops MCP Server provides an integrated interface to query and manage data from Kubernetes, Prometheus, Elasticsearch, and Jaeger through a single MCP endpoint. It exposes modules for SOPS operations, Kubernetes events, Prometheus metrics queries, Elasticsearch log searches, and Jaeger traces, allowing your AI assistants to perform operational tasks, fetch metrics, search logs, and investigate traces via natural language-driven commands. The server supports multiple protocols (HTTP/SSE and stdio-based interactions) and is configurable via YAML with environment variable overrides. To start, deploy via Docker as shown in the installation guide, then connect your MCP-enabled agent to the server URL (for example, http://<host>:80/mcp) and use the available tool commands (e.g., list-sops-from-ops, query-metrics-from-prometheus, search-logs-from-elasticsearch, get-trace-from-jaeger) to perform tasks against your observability stack.

Prerequisites:

• Docker installed and running on your host
• Access to the metrics/logs/traces backends (Kubernetes, Prometheus, Elasticsearch, Jaeger) with appropriate network connectivity

Installation steps:

1. Pull and run the Ops MCP Server Docker image (example shown in README):

   docker run -d --name ops-mcp-server
   -p 80:80
   -e SOPS_ENABLED=true
   -e EVENTS_ENABLED=true
   -e METRICS_ENABLED=true
   -e LOGS_ENABLED=true
   -e TRACES_ENABLED=true
   shaowenchen/ops-mcp-server:latest
   --mode=sse

2. Verify the server is up by hitting its health endpoint:

   curl http://localhost:80/mcp/healthz

3. Optional: configure environment-specific endpoints via a YAML config (configs/config.yaml) or pass equivalent environment variables as needed (see README for details).

4. If you prefer to run without Docker, build/run natively as per your environment once a binary or module is provided by the project maintainers (not specified in the README).

Notes and tips:

• Authentication is optional; configure SERVER_TOKEN to enable protected endpoints. Public endpoints like healthz and docs remain accessible.
• Environment variable overrides (e.g., SOPS_ENABLED, METRICS_ENABLED) can be used to tailor which modules are active at startup.
• Ensure network access from the MCP server to your backend services (Kubernetes API, Prometheus, Elasticsearch, Jaeger). Firewalls or IAM policies may require adjustments.
• If you encounter issues with specific backends, verify endpoint URLs and credentials in the YAML config or corresponding environment variables (e.g., METRICS_PROMETHEUS_, LOGS_ELASTICSEARCH_, TRACES_JAEGER_ENDPOINT).
• For production, consider using a reverse proxy, TLS termination, and rotating tokens for authentication. Monitor the server logs for startup errors and module initialization messages.