# Rove — Browser Automation API

> Zero-infrastructure Playwright API for AI agents and developers. Navigate, interact, extract structured data, and record browser sessions via REST or MCP. Token-optimized: returns accessibility trees instead of screenshots, reducing LLM token usage by ~77%.

## Quick start

- [Authentication](/docs/auth): API keys and test environment setup
- [Credits](/docs/credits): Pricing, Founder Pack, and credit deduction per action

## API reference

- [Sessions](/docs/sessions): Creating and managing browser sessions
- [Actions DSL](/docs/actions): Full command reference — navigate, click, fill, extract, screenshot
- [Artifacts](/docs/artifacts): Retrieving video recordings and screenshots
- [Errors](/docs/errors): Error codes and retry guidance

## MCP server

- [MCP setup](/docs/mcp): Install in Claude, Cursor, or VS Code in 60 seconds
- [Tool reference](/docs/mcp-tools): Full MCP tool definitions and example prompts

## Guides

- [A11y tree guide](/docs/a11y): Why accessibility trees beat screenshots for agents
- [Multi-step workflows](/docs/workflows): Maintaining session state across actions
- [Video recording](/docs/video): Debugging agent sessions with .webm recordings
- [Cost optimization](/docs/cost): Minimizing credit usage in agent workflows