Step by Step Creation of MCP Server and integration with n8n

Create an MCP server and integrate it with n8n in steps:

1. Initialize project: create a new Python virtualenv, install required packages (example: mcp-server, requests, fastapi or FastMCP).

2. Implement server: write a small FastMCP-based app that exposes REST endpoints to read and expose metrics or accept control commands. Include configuration variables for target Prometheus URL and Grafana API key.

3. Run locally: start the MCP server with uvicorn or the provided runner, verify endpoints with curl on localhost.

4. Containerize (optional): add a Dockerfile, build and push to local registry or use docker-compose with ports and environment variables.

5. Prepare n8n: run n8n in Docker or locally and open the editor.

6. Create workflow: add a Webhook or Schedule trigger node to start automation. Use an HTTP Request node to call the MCP server endpoints (GET for metrics, POST for actions). Add Function or Set nodes to transform payloads.

7. Add remediation: use SSH, Execute Command, or Docker nodes to restart services based on MCP responses. Secure calls with API keys in n8n credentials.

8. Test and monitor: trigger the workflow, inspect execution, add retries and logging.

This integration enables automated observability-driven remediation and easy orchestration between monitoring automation.

This Practical example will help in understanding the usecase