> ## Documentation Index
> Fetch the complete documentation index at: https://docs.rubixkube.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Prometheus integration

> Connect Prometheus to RubixKube today via the custom integrations path. Agents can run PromQL and pull Alertmanager state to enrich investigations and RCAs.

Prometheus connects to RubixKube today through the custom integrations path. There is no separate prebuilt app to install. Most teams start with [Custom REST](/integrations/custom-rest) pointing at the Prometheus HTTP API, and move to [Custom MCP](/integrations/custom-mcp-servers) when they want richer, semantic tools.

## What to bring in

The signals that matter most when agents are investigating or writing RCAs:

* **Instant and range queries** (`/api/v1/query`, `/api/v1/query_range`) for PromQL.
* **Series and labels** (`/api/v1/series`, `/api/v1/label/*`) for discovery.
* **Alertmanager alerts** (`/api/v2/alerts`) to surface firing alerts as insight candidates.
* **Rules and recording rules** (`/api/v1/rules`) for context on SLIs and baselines.

Works the same way with federated Prometheus, Thanos, Mimir, or any Prometheus-compatible endpoint.

## Which path to pick

<CardGroup cols={2}>
  <Card title="Start with REST" icon="network-wired" href="/integrations/custom-rest">
    Point at the Prometheus HTTP API with a small OpenAPI wrapper. Fastest path.
  </Card>

  <Card title="Move to MCP when you need more" icon="server" href="/integrations/custom-mcp-servers">
    Wrap PromQL behind named tools (for example `http_p99_latency`) so agents never see raw query strings.
  </Card>
</CardGroup>

## Quick recipe: REST path

<Steps titleSize="h3">
  <Step title="Expose Prometheus to the integration">
    Make sure the Prometheus endpoint is reachable from the RubixKube control plane. For internal-only Prometheus, use the MCP self-hosted path (below) instead.
  </Step>

  <Step title="Register the endpoint">
    In the RubixKube console, go to **Integrations → Custom REST** and register the Prometheus HTTP API (a small OpenAPI wrapper for the three or four endpoints you use).
  </Step>

  <Step title="Enable the calls you care about">
    Start with instant query, range query, and Alertmanager alerts list.
  </Step>

  <Step title="Set auth">
    If your Prometheus is protected by basic auth, bearer token, or mTLS, configure it in the integration. Otherwise no auth is needed.
  </Step>

  <Step title="Scope to skills">
    Attach the enabled endpoints to the skills that need Prometheus context.
  </Step>
</Steps>

For internal-only Prometheus behind a VPN or private network, use [Custom MCP](/integrations/custom-mcp-servers) with a self-hosted server instead, so no inbound exposure is needed.

## Example tools worth having

* `promql(query, window)` — runs a PromQL expression and returns a shaped result.
* `http_p99_latency(service, window)` — wraps a PromQL expression the team already trusts.
* `firing_alerts(service)` — returns Alertmanager alerts firing for the service.
* `recording_rule(name)` — returns the definition of a recording rule, useful when RCA wants to explain a metric.

Define these as skills (or as MCP tools if you want richer return shapes).

## Related guides

<CardGroup cols={2}>
  <Card title="Custom REST Integrations" icon="network-wired" href="/integrations/custom-rest">
    The full REST flow with auth options and approval rules.
  </Card>

  <Card title="Custom MCP Servers" icon="server" href="/integrations/custom-mcp-servers">
    When you want richer tool schemas and multi-call workflows, or when Prometheus is internal-only.
  </Card>

  <Card title="Skills" icon="sparkles" href="/concepts/skills">
    How to compose PromQL queries into custom workflows.
  </Card>
</CardGroup>