browserbase/skills
Browserbase's official collection of agent skills to access the web.
What it solves
This project provides a set of specialized capabilities (skills) that allow AI coding agents, specifically Claude Code, to interact with the web. It bridges the gap between a text-based AI agent and a live browser, enabling the agent to perform complex web tasks, automate browser interactions, and conduct deep research without manual human intervention.
How it works
It integrates with the browse CLI and Browserbase to give Claude Code a suite of tools. These tools allow the agent to launch remote or local browser sessions, handle CAPTCHAs, manage cookies, and execute serverless browser automation. The system includes specific skills for tasks like converting website traffic into API specifications, performing adversarial UI testing, and conducting automated company or competitor research.
Who it’s for
Developers and AI agent users who want to extend Claude Code's capabilities to include real-world web automation, automated QA testing, and data extraction from the web.
Highlights
- Comprehensive Browser Automation: Supports remote sessions, residential proxies, and CAPTCHAs solving.
- AI-Powered Testing: Includes adversarial UI testing that can analyze git diffs to find bugs.
- Research Automation: Specialized tools for company research, competitor analysis, and event prospecting.
- Developer Tooling: Features for syncing local Chrome cookies, capturing DevTools-protocol traces, and converting web traffic to OpenAPI specs.
- Safety Controls: A
safe-browsertool that enforces domain allowlists for agent-driven browsing.
Related
- Project
- Project
- Project
- Project
- Project