Prometheus MCP Server
An MCP server for Prometheus: the agent runs PromQL queries, discovers metrics and checks scrape targets
- 521 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Connects to Prometheus over the network and reads infrastructure metrics
- May use Prometheus access credentials
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/prometheus-mcp-serverDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Connects to Prometheus over the network and reads infrastructure metrics
- May use Prometheus access credentials
This is third-party code. Review the repository files before installing.
What it does
The server gives the agent read access to Prometheus. The agent runs instant and range PromQL queries, lists metrics with filtering and pagination, reads metric metadata and checks scrape target status. It supports basic auth, Bearer tokens, mTLS and an organization header for multi-tenant setups. The tool set can be trimmed, and a name prefix lets you connect several environments at once.
Who it is for. For DevOps and SRE engineers who investigate incidents and load using Prometheus metrics.
Good fit when
- You need to see what happened to a service during an incident
- You need a PromQL query without remembering metric names
- You need to check which scrape targets are down
Not a fit when
- Your metrics are not in a Prometheus-compatible system
- You need to change alert rules or Prometheus configuration
Example request
Show how the 5xx error rate for the checkout service changed over the last six hoursMCP
- Transport
- stdio, http, sse
- Authentication
- not required
Environment variables
PROMETHEUS_URLrequired- Prometheus server URL.
PROMETHEUS_USERNAME- Basic auth username.
PROMETHEUS_PASSWORDsecret- Basic auth password.
PROMETHEUS_TOKENsecret- Bearer token.
PROMETHEUS_MCP_SERVER_TRANSPORT- stdio, http or sse.
Limitations
Requires a Prometheus URL reachable from your machine. The README covers only the Docker image and a Helm chart, so Docker or Kubernetes is needed. Heavy queries over long ranges can load Prometheus; the default timeout is 30 seconds.
How to disable. Run claude mcp remove prometheus or delete the prometheus block from the MCP config.
FAQ
Can the agent change anything in Prometheus?
No, every tool is read-only.
How do I connect to a protected Prometheus?
Use PROMETHEUS_USERNAME and PROMETHEUS_PASSWORD, PROMETHEUS_TOKEN or mTLS certificates.
Discussion
No questions or reviews yet.
Related
MCP serversGrafana MCP Server
Grafana MCP server
Grafana's official MCP server: dashboards, Prometheus and Loki queries, alerts and incidents straight from the agent
A self-hosted gateway that merges several MCP servers into one endpoint with auth and tool selection
Datadog's official managed MCP server: logs, metrics, traces and incidents right in your agent
Self-hosted MCP server for running commands on remote servers over SSH, with tokens and roles
Zabbix MCP Server
zabbix-mcp-server
An MCP server for the whole Zabbix API: 237 tools, several Zabbix servers, scoped tokens and a read-only mode
A Go MCP server for Kubernetes and OpenShift: resources, pods, logs, Helm and Tekton straight through the cluster API
README in short
The README provides Docker-based configs for Claude Desktop, Claude Code, VS Code, Cursor and Windsurf, plus Docker Desktop and Helm chart installs. A table lists variables for URL, auth, TLS, timeout, transport and tool prefix. Six tools for queries and metric discovery are described. Python project with tests, MIT licensed.