agent-browser: Claude Code Skill & Agent Tool
An open-source skill by vercel-labs (opens in a new tab), published in the vercel-labs/agent-browser (opens in a new tab) repository on GitHub. A browser automation CLI built for AI agents: navigate, fill forms, click, take screenshots and extract data from websites — for scraping, QA and exploratory testing.
Install agent-browser
Requires Node.jsnpx skills add vercel-labs/agent-browser --skill agent-browserRun this command in your project terminal to install this skill for Claude Code.
What this skill does
agent-browser is a browser automation CLI built specifically for AI agents by Vercel Labs. It is a fast native Rust binary that drives Chrome or Chromium over the DevTools protocol, with no Playwright or Puppeteer dependency. Pages are exposed to the agent as accessibility-tree snapshots with compact element references, so it can navigate, fill forms, click buttons, take screenshots and extract data reliably while using few tokens.
The installed skill is a short discovery stub by design. Before running any command, the agent loads the real workflow guide from the CLI itself, so the instructions always match the version installed on your machine and never go stale. Specialized guides cover tasks beyond ordinary web pages: Electron desktop apps such as VS Code or Slack, Slack workspace automation, exploratory testing and bug hunts, deriving a standalone API client from a recorded HAR file, and cloud browsers.
It supports sessions, an authentication vault, state persistence and video recording, and an observability dashboard shows what the browser is doing. Install the CLI with npm and run its install command once. Use it with Claude Code or any other agent for scraping, QA and testing web applications.
Key Capabilities
- Native Rust CLI driving Chrome or Chromium over CDP, with no Playwright dependency
- Accessibility-tree snapshots with compact element refs
- Workflow guides served by the CLI, always matching the installed version
- Specialized guides for Electron apps, Slack, QA bug hunts and API client derivation
- Sessions, authentication vault, state persistence and video recording
How to use with Claude Code & AI Agents
Install the skill
Run the command above at the root of your project. The CLI asks which agents to set up — Claude Code, Cursor and other supported agents — and adds agent-browser to the project, so teammates who pull the repository get it too. Add-gto install it for your user account across every project.Open a new agent session
Start a fresh Claude Code session in that project so the agent loads the new skill. Have ready: agent-browser CLI (npm i -g agent-browser, then agent-browser install).Ask in plain words
Describe the task, for example:“Go to example.com, fill in the contact form with test data and screenshot the result.”The agent matches your request to the skill's description and follows its SKILL.md instructions step by step. Read them on GitHub before using the skill on a system that matters.