Prometheus

A Model Context Protocol (MCP) server that provides seamless integration between AI assistants andPrometheus, enabling natural language interactions with your monitoring infrastructure. This server allows for effortless querying, discovery, and analysis of metrics through Visual Studio Code, Cursor, Windsurf, Claude Desktop, and other MCP clients.

{ "mcpServers": { "prometheus": { "command": "npx", "args": ["prometheus-mcp@latest", "stdio"], "env": { "PROMETHEUS_URL": "http://localhost:9090" } } } }

# For VS Code code --add-mcp '{"name":"prometheus","command":"npx","args":["prometheus-mcp@latest","stdio"],"env":{"PROMETHEUS_URL":"http://localhost:9090"}}' # For VS Code Insiders code-insiders --add-mcp '{"name":"prometheus","command":"npx","args":["prometheus-mcp@latest","stdio"],"env":{"PROMETHEUS_URL":"http://localhost:9090"}}'

After installation, the Prometheus MCP server will be available for use with your GitHub Copilot agent in VS Code.

Go toCursor Settings→MCP→Add new MCP Server. Name to your liking, usecommandtype with the commandnpx prometheus-mcp. You can also verify config or add command arguments via clickingEdit.

{ "mcpServers": { "prometheus": { "command": "npx", "args": ["prometheus-mcp@latest", "stdio"], "env": { "PROMETHEUS_URL": "http://localhost:9090" } } } }

Follow Windsurf MCP documentation. Use the following configuration:

{ "mcpServers": { "prometheus": { "command": "npx", "args": ["prometheus-mcp@latest", "stdio"], "env": { "PROMETHEUS_URL": "http://localhost:9090" } } } }

Claude Desktop supports two installation methods:

The easiest way to install is using the pre-built DXT extension:
- Download the latest.dxtfile from thereleases page
- Double-click the downloaded file to install automatically
- Configure your Prometheus URL in the extension settings

{ "mcpServers": { "prometheus": { "command": "npx", "args": ["prometheus-mcp@latest", "stdio"], "env": { "PROMETHEUS_URL": "http://localhost:9090" } } } }

Prometheus MCP server supports the following arguments. They can be provided in the JSON configuration above, as part of the"args"list:

> npx prometheus-mcp@latest --help Commands: stdio Start Prometheus MCP server using stdio transport http Start Prometheus MCP server using HTTP transport Options: --help Show help [boolean] --version Show version number [boolean]

You can also configure the server using environment variables:

- PROMETHEUS_URL- Prometheus server URL
- ENABLE_DISCOVERY_TOOLS- Set to "false" to disable discovery tools (default: true)
- ENABLE_INFO_TOOLS- Set to "false" to disable info tools (default: true)
- ENABLE_QUERY_TOOLS- Set to "false" to disable query tools (default: true)

When running in server environments or when you need HTTP transport, run the MCP server with thehttpcommand:

npx prometheus-mcp@latest http --port 3000

And then in your MCP client config, set theurlto the HTTP endpoint:

{ "mcpServers": { "prometheus": { "command": "npx", "args": ["mcp-remote", "http://localhost:3000/mcp"] } } }

Run the Prometheus MCP server using Docker:

{ "mcpServers": { "prometheus": { "command": "docker", "args": [ "run", "-i", "--rm", "--init", "--pull=always", "-e", "PROMETHEUS_URL=http://host.docker.internal:9090", "ghcr.io/idanfishman/prometheus-mcp", "stdio" ] } } }

The Prometheus MCP server provides 10 tools organized into three configurable categories:

Tools for exploring your Prometheus infrastructure:

- Description: List all available Prometheus metrics
- Parameters: None
- Read-only:true

- Description: Get metadata for a specific Prometheus metric
- Parameters:

- metric(string): Metric name to get metadata for

- Description: List all available Prometheus labels
- Parameters: None
- Read-only:true

- Description: Get all values for a specific Prometheus label
- Parameters:

- label(string): Label name to get values for

- Description: List all Prometheus scrape targets
- Parameters: None
- Read-only:true

- Description: Get targets for a specific scrape pool
- Parameters:

- scrapePool(string): Scrape pool name

Tools for accessing Prometheus server information:

- Description: Get Prometheus runtime information
- Parameters: None
- Read-only:true

- Description: Get Prometheus build information
- Parameters: None
- Read-only:true

- Description: Execute an instant Prometheus query
- Parameters:

- query(string): Prometheus query expression
- time(string, optional): Time parameter for the query (RFC3339 format)

- Description: Execute a Prometheus range query
- Parameters:

- query(string): Prometheus query expression
- start(string): Start timestamp (RFC3339 or unix timestamp)
- end(string): End timestamp (RFC3339 or unix timestamp)
- step(string): Query resolution step width

Here are some example interactions you can have with your AI assistant:

- "Show me all available metrics in Prometheus"
- "What's the current CPU usage across all instances?"
- "Get the memory usage for the last hour"

- "List all scrape targets and their status"
- "What labels are available for thehttp_requests_totalmetric?"
- "Show me all metrics related to 'cpu'"

- "Compare CPU usage between production and staging environments"
- "Show me the top 10 services by memory consumption"
- "What's the error rate trend for the API service over the last 24 hours?"

- Network Access: The server requires network access to your Prometheus instance
- Resource Usage: Range queries can be resource-intensive; monitor your Prometheus server load

- Verify your Prometheus server is accessible at the configured URL
- Check firewall settings and network connectivity
- Ensure Prometheus API is enabled (default on port 9090)

- Verify the MCP server has network access to Prometheus
- Check if authentication is required for your Prometheus setup

- If certain tools are missing, check if they've been disabled via configuration

This project is licensed under the MIT License - see theLICENSEfile for details.

- GitHub Issues:Report bugs or request features
- Documentation:Model Context Protocol Documentation
- Prometheus:Prometheus Documentation

Expose data observability, lineage, test results & incidents to AI agents via MCP