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

# Claude Scope: AI Prompt Generator for UI Debugging

> Claude Scope records your browser tab, runs Vision AI and Playwright analysis, then generates a structured prompt for Claude Code, Codex, or Cursor.

Claude Scope is a developer tool that bridges the gap between what you see in your browser and what your AI coding agent needs to fix it. Instead of writing lengthy bug descriptions by hand, you record your screen, and Claude Scope automatically produces a structured system prompt — complete with a visual timeline, an ARIA accessibility tree, and a DOM diff — ready to drop into Claude Code, Codex, Cursor, or any agent that accepts a raw prompt.

## What Claude Scope does

When you start a recording, Claude Scope captures your browser tab as a WebM video and runs it through a dual-lane analysis pipeline. The **Vision lane** uses Anthropic's Vision AI to analyze each extracted frame and identify UI elements: buttons, inputs, headings, links, and their states. The **Playwright lane** launches a headless browser against your seed URL and captures a full ARIA accessibility snapshot, counting every interactive element by role and name. Both results are then synthesized into a single, formatted system prompt targeted at the agent you choose.

<Note>
  Claude Scope uses the BYOK (bring your own key) model — you supply your own Anthropic API key for vision analysis. Claude Scope itself is free during the beta period; you only pay Anthropic directly for the API calls you make.
</Note>

## Key capabilities

| Capability        | Details                                                                        |
| ----------------- | ------------------------------------------------------------------------------ |
| Screen recording  | Captures any browser tab using the MediaRecorder API (WebM/VP9)                |
| Frame extraction  | SSIM-based differencing extracts only frames where the UI meaningfully changed |
| Vision lane       | Anthropic Vision AI identifies UI elements frame-by-frame                      |
| Playwright lane   | Headless browser captures a full ARIA accessibility snapshot                   |
| Prompt generation | Both lanes merge into one structured, agent-ready system prompt                |
| Agent targets     | Claude Code, Codex, Cursor, and Raw (unformatted)                              |

## Supported AI agents

Claude Scope formats the generated prompt differently depending on your target agent:

* **Claude Code** — Structured system prompt with inline ARIA tree and screenshot bundle
* **Codex** — Compact diff-focused prompt suited for GPT-4o-based completions
* **Cursor** — Prompt formatted for Cursor's composer context window
* **Raw** — Unformatted output for pasting into any other tool or LLM

<Tip>
  You can change the agent target before copying your prompt. The underlying analysis is the same — only the formatting changes.
</Tip>

## Pricing

Claude Scope is free during beta. All features — unlimited recordings, dual-lane analysis, and prompt generation for every supported agent — are included at no cost. You only pay Anthropic for API usage against your own key.

## Where to go next

<CardGroup cols={2}>
  <Card title="Quickstart" icon="bolt" href="/quickstart">
    Sign up, add your API key, and run your first recording in under five minutes.
  </Card>

  <Card title="How it works" icon="microchip" href="/how-it-works">
    Understand the full pipeline from video capture to generated prompt.
  </Card>

  <Card title="New recording" icon="circle-dot" href="/recording/new-recording">
    Step-by-step guide to setting up and running a recording session.
  </Card>

  <Card title="Model Access" icon="key" href="/configuration/model-access">
    Configure your Anthropic API key and choose a model for vision analysis.
  </Card>
</CardGroup>