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

# AI SRE Documentation

> AI-powered incident response assistant

<Hero title="AI SRE" subtitle="Investigate and resolve production incidents using AI-powered analysis of your code, infrastructure, and telemetry." image="/logo.svg" />

<CardGroup cols={2}>
  <Card title="Get Started" icon="rocket" href="/getting-started/overview">
    Set up AI SRE
  </Card>

  <Card title="Overview" icon="lightbulb" href="/working-with-ai-sre/overview">
    How AI SRE works
  </Card>

  <Card title="Workflows" icon="wrench" href="/working-with-ai-sre/overview/detection">
    Detection, triage, RCA
  </Card>

  <Card title="Connectors" icon="plug" href="/connectors/observability">
    Connect your tools
  </Card>
</CardGroup>

## What AI SRE does

AI SRE investigates production incidents by accessing your code repositories, infrastructure, and telemetry. It does the heavy lifting: querying systems, correlating data, building evidence chains, and identifying root causes—work that typically takes hours.

**Incident investigation** — Reviews code changes, determines scope, queries multiple systems, and builds evidence chains to identify root causes.

**Change correlation** — Links recent deployments, commits, and configuration changes to incidents.

**Impact assessment** — Evaluates blast radius, quantifies business impact, and identifies affected services and users.

## How it works

1. **Connect your stack** — Link your code repositories, telemetry, infrastructure, knowledge base, and alerting systems
2. **Ask questions** — Describe incidents in natural language
3. **AI SRE investigates** — Accesses code repositories, queries logs and metrics, correlates changes, and builds evidence chains across your systems
4. **Review findings** — Get evidence-based conclusions with confidence levels and suggested actions
5. **Resolve faster** — Use insights to fix incidents in minutes instead of hours