dev-browser
Let your AI agent use a browser through sandboxed Playwright scripts
Description
dev-browser gives AI agents and developers the ability to control browsers through sandboxed JavaScript execution. Scripts run in a QuickJS WASM sandbox with no host access, using the full Playwright API for navigation, clicking, filling forms, screenshots, and more. Pages persist across scripts, and it auto-connects to running Chrome or launches fresh Chromium. Designed to be the fastest way for an agent to use a browser by letting it write code.
Install
npm install -g dev-browserAI Summary
Browser automation CLI for AI agents — sandboxed Playwright scripts with persistent pages, auto-connect to Chrome, and AI-friendly page snapshots.
Capabilities
- + Control browsers via sandboxed JavaScript in QuickJS WASM
- + Full Playwright API: goto, click, fill, locators, evaluate, screenshots
- + Persistent page sessions across multiple scripts
- + Auto-connect to running Chrome or launch fresh Chromium
- + AI-friendly page snapshots via snapshotForAI()
- + Headless and headed modes
- + Sandboxed file I/O restricted to ~/.dev-browser/tmp/
Use When
- → You want an AI agent to interact with web pages
- → You need browser automation with sandboxed script execution
- → You want to let an agent write Playwright code to browse the web
Avoid When
- x You only need to fetch static page content (use curl or firecrawl)
- x You need browser automation without Node.js
Usage Patterns
Install and set up
npm i -g dev-browser && dev-browser install Install globally and download Chromium
Run headless
dev-browser --headless Launch a headless browser session for scripting
Connect to running Chrome
dev-browser --connect Attach to an already running Chrome instance
Input / Output
Typical Pipelines
dev-browser --headless < script.js