prometheus-mcp-server

MCP.Pizza Chef: pab1it0

Prometheus collects numbers about how your servers and apps are behaving: traffic, errors, memory use, response times. With this in place you can ask what happened at three in the morning instead of writing a query by hand, and get hours of history summarised in a sentence. It runs as a Docker container and works in Claude Desktop, Claude Code, VS Code, Cursor and Windsurf. You point it at your own Prometheus address; a username, password or token is only needed if yours demands one.

Coding

Data

Use This MCP server To

Ask what caused the spike in errors last night Compare this week's response times against last week List which services are currently being monitored Check whether a machine has stopped reporting in Summarise memory use across the last twenty-four hours

README

Prometheus MCP Server

Python

Give AI assistants the power to query your Prometheus metrics.

A Model Context Protocol (MCP) server that provides access to your Prometheus metrics and queries through standardized MCP interfaces, allowing AI assistants to execute PromQL queries and analyze your metrics data.

Getting Started

Prerequisites

Prometheus server accessible from your environment
MCP-compatible client (Claude Desktop, VS Code, Cursor, Windsurf, etc.)

Installation Methods

prometheus-mcp-server FAQ

Is this project still maintained?

Yes, and actively. It was updated within the last few days and ships tagged releases regularly.

Do I need my own Prometheus setup?

Yes. This reads from a Prometheus server you already run; it does not collect anything itself.

Do I need a key or password?

Only if your Prometheus is protected. It supports a username and password, a token, or client certificates, and works with none of them if yours is open.

Which apps does it work in?

Claude Desktop, Claude Code, VS Code, Cursor and Windsurf, plus Docker Desktop's own catalogue.

Can I use this to work out why a service was slow yesterday?

Yes. You can ask for a range of time and it pulls the matching history back, including which machines were affected.

How hard is setup?

You need Docker running, then a small block of settings pointing at your Prometheus address. There is also a Kubernetes chart if you deploy that way.