Prometheus

You can access Prometheus data through a Model Context Protocol (MCP) server that acts as a bridge between Prometheus and MCP clients. This server makes it easy to query Prometheus metrics from MCP-enabled tools and dashboards, enabling streamlined integration and multi-tenant access patterns.

How to use

Connect to a Prometheus MCP Server from your MCP client by using the server URL and, if needed, authentication credentials. You can run the MCP server locally or point your client to a remote MCP endpoint. Use the client to query metrics, organize data into models, and drive dashboards or automations that rely on Prometheus data. If you operate in a multi-tenant environment, the server supports per-tenant configuration through the provided authentication options.

How to install

Prerequisites: you need Python and a package runner. You will install and run the MCP server client that exposes Prometheus data through MCP.

pipx install git+https://github.com/moohoorama/prometheus-mcp-server-py.git

If you prefer not to install the package, you can run it on demand using a one-time execution command.

pipx run --spec git+https://github.com/moohoorama/prometheus-mcp-server-py.git prometheus-mcp --url http://your-prometheus-server:9090

Additional notes

Authentication options include a username/password combination or an access token. When connecting, provide the Prometheus URL with the --url flag and supply credentials if your Prometheus instance requires Basic Authentication. If your deployment uses a token-based approach, you can supply --token instead of username and password. The MCP server supports these authentication methods to fit into secure environments and multi-tenant setups.