Prometheus

You run a Prometheus MCP Server to enable large language models to interact with a Prometheus instance through the API. It lets you generate and execute promql queries, inspect metrics, and leverage Prometheus capabilities directly from your MCP-enabled workflows.

How to use

You start the MCP server and connect an MCP client to issue Prometheus API calls. The server exposes tools that let you query metrics, inspect targets and rules, read documentation, and more. You can run the server in different transports (stdio or HTTP) and pass Prometheus connection details via flags or environment variables. Use core tools by default for efficient metric discovery, and add more tools as needed.

How to install

Prerequisites: you need a working Go environment for building from source, Docker for container-based runs, and access to a Prometheus instance you want to query.

Option A: Run with Docker (stdio transport) using a container image.

Option B: Run with Docker (streamable HTTP transport) to expose HTTP endpoints.

Follow the exact commands below to start the MCP server with a Prometheus URL. Replace the placeholder with your Prometheus URL.

Additional notes and configuration

The MCP server supports a broad set of tools to interact with Prometheus. By default, a core set of tools is loaded to provide essential capabilities. You can enable additional tools as needed. You can also pass in environment variables to control the Prometheus connection and server behavior.

Telemetry and security options are available. The server exposes metrics at the configured listen address and path, and you can secure access using TLS or authentication by providing a web configuration file.

Troubleshooting

If you encounter connection issues to Prometheus, verify the Prometheus URL is reachable from the MCP host and that any required TLS or authentication settings are provided via config options or environment variables.

If the MCP server reports missing tools or endpoints, confirm you loaded the desired tool set (core tools are loaded by default) and that you are using a Prometheus-compatible backend that exposes the expected API endpoints.

Available tools

alertmanagers

Get overview of Prometheus Alertmanager discovery

build_info

Get Prometheus build information

config

Get Prometheus configuration

docs_list

List of Official Prometheus Documentation Files

docs_read

Read a named markdown file from the official Prometheus docs

docs_search

Search Prometheus documentation files

exemplars_query

Query exemplars for a given time range

flags

Get runtime flags of the Prometheus server

healthy

Health check endpoint for Prometheus API

label_names

Return unique label names for a time range and matchers

label_values

Query values for a specific label over a time range

list_alerts

List all active alerts

list_rules

List all loaded alerting and recording rules

list_targets

Overview of target discovery

metric_metadata

Return metadata about metrics being scraped

query

Execute an instant query against Prometheus

quit

Graceful shutdown of Prometheus

range_query

Execute a range query against Prometheus

ready

Check if Prometheus is ready to serve traffic

reload

Reload Prometheus configuration and rules

runtime_info

Get Prometheus runtime information

series

Find series by label matchers

targets_metadata

Metadata about metrics scraped by targets

tsdb_stats

TSDB usage and cardinality statistics

wal_replay_status

Current WAL replay status

clean_tombstones

Remove deleted data and tombstones from disk

delete_series

Delete data for a selection of series in a time range

snapshot

Create a snapshot of current data in the TSDB