Prometheus

You can connect Claude or other MCP clients to your Prometheus data through this MCP server, enabling you to list Prometheus metrics, access detailed metric metadata, and retrieve current statistics through a stable MCP interface.

How to use

Start the MCP Prometheus bridge and point your client at it to browse metrics, view metadata, and fetch current statistics. You will list available metrics, read detailed information including help text and type, and obtain recent statistical data such as count, min, and max for each metric. Use the provided environment variables to configure access to your Prometheus instance, including optional basic authentication if your Prometheus server requires it.

How to install

Prerequisites: you need Node.js installed on your machine. You will also run a build step to produce the executable that the MCP server uses.

Install dependencies and build the server from source.

Additional configuration and usage notes

Configuration revolves around pointing the MCP server at your Prometheus instance and optionally enabling basic authentication. The following environment variables are used by the server.

Configuration details

{
  "mcpServers": {
    "prometheus": {
      "command": "/path/to/mcp-server-prometheus/build/index.js",
      "env": {
        "PROMETHEUS_URL": "http://your-prometheus-instance:9090",
        "PROMETHEUS_USERNAME": "YOUR_USERNAME",
        "PROMETHEUS_PASSWORD": "YOUR_PASSWORD"
      }
    }
  }
}

Security and troubleshooting

If you enable basic authentication, ensure your credentials are kept secure and not exposed in logs or shared configurations. If you encounter issues connecting to Prometheus, verify that PROMETHEUS_URL is reachable from the MCP server host and that any firewall rules allow traffic to Prometheus.

For debugging, you can run the MCP server in development mode to get live feedback during startup and interaction with Prometheus.

Available tools

list_metrics

Lists all available Prometheus metrics with their descriptions and basic metadata.

read_metric

Reads detailed information about a specific metric, including help text, type, and metadata.

read_stats

Retrieves current statistical information for a metric, such as count, min, and max values.