This project provides a browser automation skill for Claude Code, enabling AI agents to interact with web browsers via natural language. It targets users of Claude Code who need to automate web tasks, offering a more intuitive interface than traditional browser automation frameworks by leveraging AI for instruction interpretation. The primary benefit is simplifying complex web interactions for AI agents.

How It Works

The skill integrates Claude Code with the Stagehand AI browser automation framework. Stagehand accepts natural language instructions, translating them into browser actions. This approach is designed to be more context-efficient than native tools like Playwright, while offering specialized features for AI-driven automation tasks. The agent interprets user requests and executes them through the browser interface.

Quick Start & Requirements

• Installation:
  • Add marketplace: bash /plugin marketplace add browserbase/agent-browse
  • Install plugin: bash /plugin install browser-automation@browser-tools
  • (Manual installation steps are also provided in the README).
  • Restart Claude Code after installation.
• Prerequisites:
  • Google Chrome installed on the user's platform (macOS, Windows, Linux).
  • An Anthropic API key set as an environment variable: export ANTHROPIC_API_KEY="your-api-key".
• Resources:
  • Stagehand Documentation: [Link not provided in README, assumed internal or separate]
  • Claude Code Skills: [Link not provided in README]

Highlighted Details

• Enables AI agents to perform complex web tasks like data extraction, summarization, and e-commerce interactions using natural language prompts.
• Leverages Stagehand for AI-powered browser automation, aiming for higher context efficiency compared to Playwright.
• Supports direct interaction with websites, including logged-in sessions and QA testing scenarios.

Maintenance & Community

Information regarding maintainers, community channels (like Discord/Slack), project roadmap, or sponsorships is not detailed in the provided README.

Licensing & Compatibility

The license type and any compatibility notes for commercial use or integration with closed-source projects are not specified in the README.

Limitations & Caveats

The primary requirement is the installation of Google Chrome. Troubleshooting involves ensuring Chrome is correctly installed and accessible. The README also mentions a method for refreshing the Chrome profile (rm -rf .chrome-profile), suggesting potential issues with persistent state or cookies. The project appears to be a skill for Claude Code, implying it may not be a standalone tool.