Creativity · MCP — server

MCP Prometheus Server

The MCP Prometheus Server is a community-maintained Model Context Protocol server that wraps Prometheus's HTTP API. It lets an LLM client run instant and range PromQL queries, list label values, discover series, and inspect target health — useful for incident triage and capacity analysis from chat.

MCP facts

Kind: server
Ecosystem: anthropic-mcp
Language: Python
Transports: stdio

Capabilities

Tools: query, query_range, series, label_values, targets
Resources: prom://series/{metric_name} URIs
Auth: optional PROMETHEUS_BEARER_TOKEN or basic auth

Install

pip install mcp-server-prometheus && python -m mcp_server_prometheus

Configuration

{
  "mcpServers": {
    "prometheus": {
      "command": "python",
      "args": ["-m", "mcp_server_prometheus"],
      "env": {
        "PROMETHEUS_URL": "https://prom.example.com",
        "PROMETHEUS_BEARER_TOKEN": "optional-token"
      }
    }
  }
}

Frequently asked questions

Is this server official?

No — it is community-maintained. The Prometheus project has not released a first-party MCP server.

Can the model write alerts?

Not through Prometheus itself — it is read-only. For alert management, pair this server with an MCP server for Alertmanager or your incident tool.

How do I limit query cost?

Set short default step sizes for range queries, cap timestamps at a max lookback, and block series matchers without a metric name to prevent expensive scans.

Sources

Prometheus HTTP API — accessed 2026-04-20
MCP community servers — accessed 2026-04-20