Yes. prometheus-mcp-server is free and open source under the MIT license. You only need access to a Prometheus-compatible endpoint; there is no separate fee for the server.

Does it support remote/OAuth?

No. It runs locally over stdio (via the Docker image or the PyPI package) and connects to your Prometheus HTTP API. It supports basic-auth, a bearer token, and custom headers rather than an interactive OAuth flow.

Does it work with Mimir, Cortex, or Thanos?

Yes. It targets the Prometheus HTTP API, so any compatible backend works; set ORG_ID for multi-tenant setups and PROMETHEUS_TOKEN or basic-auth credentials as your gateway requires.

Prometheus MCP server

Communitypab1it0450Config last verified Jun 1, 2026

A maintained Prometheus MCP server: run PromQL instant and range queries, discover metrics and metadata, and inspect scrape targets from your agent.

prometheus-mcp-server is an actively maintained Model Context Protocol server that lets an AI client query and analyze a Prometheus instance directly. It hands the agent the core observability primitives: execute PromQL instant queries for a point-in-time value, run range queries over a start/end window with a step to chart a trend, list the available metrics with pagination and filtering, fetch metadata for one metric or in bulk, and inspect all scrape targets to see what is up and being collected. A health-check tool rounds it out for container monitoring. Together these let an agent investigate a spike, confirm which series exist before writing a query, and check whether a target is reporting — without you hand-writing every PromQL expression.

It connects to any Prometheus-compatible HTTP API via the required PROMETHEUS_URL, and supports basic-auth (PROMETHEUS_USERNAME / PROMETHEUS_PASSWORD), bearer-token (PROMETHEUS_TOKEN), multi-tenant ORG_ID, custom headers, and request timeouts. The canonical way to run it is the published Docker image ghcr.io/pab1it0/prometheus-mcp-server over stdio; it is also published to PyPI as prometheus-mcp-server. The project is open source under the MIT license.

Quick install

Copy-paste configs are provided for all 8 supported clients. Pick your client below.

Add to ~/.claude.json

~/.claude.json

json

{
  "mcpServers": {
    "prometheus": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "-e",
        "PROMETHEUS_URL",
        "ghcr.io/pab1it0/prometheus-mcp-server:latest"
      ],
      "env": {
        "PROMETHEUS_URL": "<PROMETHEUS_URL>",
        "PROMETHEUS_USERNAME": "<PROMETHEUS_USERNAME>",
        "PROMETHEUS_PASSWORD": "<PROMETHEUS_PASSWORD>",
        "PROMETHEUS_TOKEN": "<PROMETHEUS_TOKEN>",
        "ORG_ID": "<ORG_ID>"
      }
    }
  }
}

Or via CLI

bash

claude mcp add prometheus -- docker run -i --rm -e PROMETHEUS_URL ghcr.io/pab1it0/prometheus-mcp-server:latest

Step-by-step guides

Available tools

Tool	Description
health_check	Health-check endpoint for container monitoring and status verification.
execute_query	Executes a PromQL instant query against Prometheus.
execute_range_query	Executes a PromQL range query with start time, end time, and step interval.
list_metrics	Lists all available metrics in Prometheus, with pagination and filtering support.
get_metric_metadata	Gets metadata for one metric, or bulk metadata with optional filtering.
get_targets	Gets information about all scrape targets.

Required configuration

PROMETHEUS_URLRequired
Base URL of your Prometheus-compatible HTTP API, e.g. http://localhost:9090.
PROMETHEUS_USERNAMEOptional
Username for HTTP basic auth. Optional.
PROMETHEUS_PASSWORDOptional
Password for HTTP basic auth. Optional.
PROMETHEUS_TOKENOptional
Bearer token for authentication. Optional.
ORG_IDOptional
Organization/tenant ID header for multi-tenant setups. Optional.
PROMETHEUS_REQUEST_TIMEOUTOptional
Request timeout in seconds for Prometheus API calls. Optional.
PROMETHEUS_CUSTOM_HEADERSOptional
Custom headers to send with every Prometheus request. Optional.

What you can do with it

Investigate a metric spike

The agent runs a range query over the window around an incident to chart the trend, then drills into related series with instant queries to find the dimension — pod, instance, or status code — driving the spike.

Write correct PromQL without guessing

Before composing a query the agent lists available metrics and pulls metric metadata to confirm names, types, and help text, then checks scrape targets to verify the data is actually being collected.

FAQ

Is it free?: Yes. prometheus-mcp-server is free and open source under the MIT license. You only need access to a Prometheus-compatible endpoint; there is no separate fee for the server.
Does it support remote/OAuth?: No. It runs locally over stdio (via the Docker image or the PyPI package) and connects to your Prometheus HTTP API. It supports basic-auth, a bearer token, and custom headers rather than an interactive OAuth flow.
Does it work with Mimir, Cortex, or Thanos?: Yes. It targets the Prometheus HTTP API, so any compatible backend works; set ORG_ID for multi-tenant setups and PROMETHEUS_TOKEN or basic-auth credentials as your gateway requires.

← Browse all monitoring-observability servers