Query Prometheus Metrics with AI
MCP server letting AI assistants run PromQL queries and explore metrics, targets, and metadata against Prometheus.
Why it matters
Enable AI assistants to directly query and analyze your Prometheus monitoring data. Execute PromQL queries, explore metrics, and gain insights into system health and performance.
Outcomes
What it gets done
Execute instant and range PromQL queries.
List and retrieve metadata for Prometheus metrics.
Access data collection target information.
Authenticate with Prometheus using basic auth or bearer tokens.
Install
Add it to your toolbox
Run in your project directory:
curl -fsSL https://spark.entire.vc/get/vb-prometheus | bash Capabilities
Tools your agent gets
Health check for container monitoring and status verification
Execute an instant PromQL query against Prometheus
Execute a range PromQL query with start time, end time, and step interval
List all available metrics in Prometheus with pagination and filtering support
Get metadata for a specific metric
Get information about all data collection targets
Overview
Prometheus MCP Server
An MCP server exposing PromQL query execution, metric discovery, and scrape target inspection against a Prometheus server, with configurable tool selection and multiple auth methods. Use to let an AI assistant query and explore Prometheus metrics directly. Requires a reachable Prometheus server and, for secured instances, basic/bearer/mTLS auth configured.
What it does
Prometheus MCP Server gives AI assistants direct access to a Prometheus server through standardized MCP interfaces, letting them execute PromQL queries and analyze metrics data. It exposes six tools: health_check (container monitoring/status), execute_query (run a PromQL instant query), execute_range_query (run a PromQL range query with start time, end time, and step interval), list_metrics (list available metrics with pagination and filtering), get_metric_metadata (get metadata for one metric or in bulk with filtering), and get_targets (information about all scrape targets). The tool list is configurable per deployment, so unused tools can be excluded to save context tokens.
docker run -i --rm \
-e PROMETHEUS_URL="http://your-prometheus:9090" \
ghcr.io/pab1it0/prometheus-mcp-server:latest
Configuration is entirely environment-variable driven: PROMETHEUS_URL is the only required setting; optional variables cover SSL verification toggling, disabling Prometheus UI links in results to save tokens, request timeout (default 30s, guards against hanging requests), basic auth (username/password), bearer token auth, mutual TLS via client cert/key, a CA bundle path, multi-tenant ORG_ID, transport mode (stdio/http/sse, default stdio), HTTP bind host/port, stateless HTTP mode for multi-replica deployments, custom headers as JSON, and a TOOL_PREFIX for running multiple instances against different environments (e.g. "staging" produces staging_execute_query) side by side in the same client.
Features include executing PromQL instant and range queries, discovering and exploring metrics (listing available metrics, getting per-metric metadata, searching metadata by name or description in a single call), authentication support (basic auth and bearer token from environment variables), and Docker containerization.
When to use - and when NOT to
Use this connector when you want an AI assistant to query and explore Prometheus metrics directly - running ad hoc PromQL queries, checking scrape target health, discovering what metrics are available, or pulling range-query data for trend analysis, without leaving the conversation to open a Prometheus dashboard.
It requires a reachable Prometheus server (PROMETHEUS_URL) and, for anything beyond an open/unauthenticated instance, one of the supported auth methods configured. Because tool selection is configurable, deployments that only need querying (not target discovery, for example) can trim the exposed tool set to reduce context usage.
Capabilities
Query: execute_query (instant PromQL), execute_range_query (range PromQL with step interval). Discovery: list_metrics (paginated, filterable), get_metric_metadata (single or bulk), get_targets (scrape target info). System: health_check for monitoring.
How to install
Run via Docker with PROMETHEUS_URL set (and PROMETHEUS_USERNAME/PROMETHEUS_PASSWORD for authenticated instances), configured directly in Claude Desktop, Claude Code (via claude mcp add prometheus), VS Code, Cursor, or Windsurf's MCP settings. Also installable through Docker Desktop's MCP Toolkit/Catalog, or deployed to Kubernetes via the published Helm chart (helm install prometheus-mcp-server oci://ghcr.io/pab1it0/charts/prometheus-mcp-server), with authentication and custom values file options supported.
Who it's for
SREs, DevOps engineers, and platform teams who want an AI assistant to query and explore Prometheus metrics and scrape targets directly during incident response or routine monitoring work.
Source README
Prometheus MCP Server
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
Claude Desktop
Add to your Claude Desktop configuration:
{
"mcpServers": {
"prometheus": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"PROMETHEUS_URL",
"ghcr.io/pab1it0/prometheus-mcp-server:latest"
],
"env": {
"PROMETHEUS_URL": "<your-prometheus-url>"
}
}
}
}
Claude Code
Install via the Claude Code CLI:
claude mcp add prometheus --env PROMETHEUS_URL=http://your-prometheus:9090 -- docker run -i --rm -e PROMETHEUS_URL ghcr.io/pab1it0/prometheus-mcp-server:latest
VS Code / Cursor / Windsurf
Add to your MCP settings in the respective IDE:
{
"prometheus": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"PROMETHEUS_URL",
"ghcr.io/pab1it0/prometheus-mcp-server:latest"
],
"env": {
"PROMETHEUS_URL": "<your-prometheus-url>"
}
}
}
Docker Desktop
The easiest way to run the Prometheus MCP server is through Docker Desktop:
Via MCP Catalog: Visit the Prometheus MCP Server on Docker Hub and click the button above
Via MCP Toolkit: Use Docker Desktop's MCP Toolkit extension to discover and install the server
Configure your connection using environment variables (see Configuration Options below)
Manual Docker Setup
Run directly with Docker:
# With environment variables
docker run -i --rm \
-e PROMETHEUS_URL="http://your-prometheus:9090" \
ghcr.io/pab1it0/prometheus-mcp-server:latest
# With authentication
docker run -i --rm \
-e PROMETHEUS_URL="http://your-prometheus:9090" \
-e PROMETHEUS_USERNAME="admin" \
-e PROMETHEUS_PASSWORD="password" \
ghcr.io/pab1it0/prometheus-mcp-server:latest
Helm Chart (Kubernetes)
Deploy to Kubernetes using the Helm chart from the OCI registry:
helm install prometheus-mcp-server \
oci://ghcr.io/pab1it0/charts/prometheus-mcp-server \
--version 1.0.0 \
--set prometheus.url="http://prometheus:9090"
With authentication:
helm install prometheus-mcp-server \
oci://ghcr.io/pab1it0/charts/prometheus-mcp-server \
--version 1.0.0 \
--set prometheus.url="http://prometheus:9090" \
--set auth.username="admin" \
--set auth.password="secret"
With a custom values file:
helm install prometheus-mcp-server \
oci://ghcr.io/pab1it0/charts/prometheus-mcp-server \
--version 1.0.0 \
-f values.yaml
See the chart values for all available configuration options.
Configuration Options
| Variable | Description | Required |
|---|---|---|
PROMETHEUS_URL |
URL of your Prometheus server | Yes |
PROMETHEUS_URL_SSL_VERIFY |
Set to False to disable SSL verification | No |
PROMETHEUS_DISABLE_LINKS |
Set to True to disable Prometheus UI links in query results (saves context tokens) | No |
PROMETHEUS_REQUEST_TIMEOUT |
Request timeout in seconds to prevent hanging requests (DDoS protection) | No (default: 30) |
PROMETHEUS_USERNAME |
Username for basic authentication | No |
PROMETHEUS_PASSWORD |
Password for basic authentication | No |
PROMETHEUS_TOKEN |
Bearer token for authentication | No |
PROMETHEUS_CLIENT_CERT |
Path to client certificate file for mutual TLS authentication | No |
PROMETHEUS_CLIENT_KEY |
Path to client private key file for mutual TLS authentication | No |
REQUESTS_CA_BUNDLE |
Path to CA bundle file for verifying the server's TLS certificate (standard requests library env var) |
No |
ORG_ID |
Organization ID for multi-tenant setups | No |
PROMETHEUS_MCP_SERVER_TRANSPORT |
Transport mode (stdio, http, sse) | No (default: stdio) |
PROMETHEUS_MCP_BIND_HOST |
Host for HTTP transport | No (default: 127.0.0.1) |
PROMETHEUS_MCP_BIND_PORT |
Port for HTTP transport | No (default: 8080) |
PROMETHEUS_MCP_STATELESS_HTTP |
Enable stateless HTTP mode for multi-replica support | No (default: False) |
PROMETHEUS_CUSTOM_HEADERS |
Custom headers as JSON string | No |
TOOL_PREFIX |
Prefix for all tool names (e.g., staging results in staging_execute_query). Useful for running multiple instances targeting different environments in Cursor |
No |
Available Tools
| Tool | Category | Description |
|---|---|---|
health_check |
System | Health check endpoint for container monitoring and status verification |
execute_query |
Query | Execute a PromQL instant query against Prometheus |
execute_range_query |
Query | Execute a PromQL range query with start time, end time, and step interval |
list_metrics |
Discovery | List all available metrics in Prometheus with pagination and filtering support |
get_metric_metadata |
Discovery | Get metadata for one metric or bulk metadata with optional filtering |
get_targets |
Discovery | Get information about all scrape targets |
The list of tools is configurable, so you can choose which tools you want to make available to the MCP client. This is useful if you don't use certain functionality or if you don't want to take up too much of the context window.
Features
- Execute PromQL queries against Prometheus
- Discover and explore metrics
- List available metrics
- Get metadata for specific metrics
- Search metric metadata by name or description in a single call
- View instant query results
- View range query results with different step intervals
- Authentication support
- Basic auth from environment variables
- Bearer token auth from environment variables
- Docker containerization support
- Provide interactive tools for AI assistants
Development
Contributions are welcome! Please see our Contributing Guide for detailed information on how to get started, coding standards, and the pull request process.
This project uses uv to manage dependencies. Install uv following the instructions for your platform:
curl -LsSf https://astral.sh/uv/install.sh | sh
You can then create a virtual environment and install the dependencies with:
uv venv
source .venv/bin/activate # On Unix/macOS
.venv\Scripts\activate # On Windows
uv pip install -e .
Testing
The project includes a comprehensive test suite that ensures functionality and helps prevent regressions.
Run the tests with pytest:
# Install development dependencies
uv pip install -e ".[dev]"
# Run the tests
pytest
# Run with coverage report
pytest --cov=src --cov-report=term-missing
When adding new features, please also add corresponding tests.
FAQ
Common questions
Discussion
Questions & comments · 0
Sign In Sign in to leave a comment.