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

# Bench

> Find AI failures, investigate the evidence, and review improvements.

Bench connects your prompts, code and business context to repeatable evaluations. Start with a repository or uploaded prompts. Inspect discovered AI systems, run a bench, and review the failed checks and proposed improvements.

<Info>
  These guides include the system-first staging preview. SDK monitoring, system-wide Free allowances and the new context tools require the matching preview deployment. They are not a production release announcement. Use the setup instructions and endpoint shown inside your Bench environment.
</Info>

<CardGroup cols={2}>
  <Card title="Run your first bench" icon="play" href="/quickstart">Connect a source and evaluate a prompt.</Card>
  <Card title="Connect your coding agent" icon="terminal" href="/mcp/overview">Use Claude Code, Cursor or Codex.</Card>
  <Card title="Define success" icon="bullseye" href="/guides/context">Add business rules and examples when useful.</Card>
  <Card title="Build a test library" icon="table" href="/guides/test-library">Keep trusted cases and customize criteria.</Card>
</CardGroup>

## What a bench does

Bench creates criteria and scenarios, measures the current prompt, then tests candidate prompt and model changes on that suite. You can inspect the cases, scores, explanations and comparison before accepting a recommendation.

<Note>
  A code scan is not a runtime evaluation. Current prompt-level evaluations do not execute an entire application, agent graph or production tool side effect. The result page distinguishes a suggested improvement from an applied change.
</Note>