Playwright MCP
Lets your AI agent open web pages, click, type and fill in forms in a real browser
by Microsoft
Dev Browser
by Sawyer HoodCLI toolWrite code
Lets your coding agent drive a real Chrome browser to open pages, click and check results
About Dev Browser
Dev Browser is a command-line tool that lets an AI coding agent drive Chrome: open pages, click, fill in forms and take screenshots. It can start its own browser or attach to one you already have open.
For People building with coding agentsDevelopers testing web apps
Made by Sawyer Hood
An open-source project by developer Sawyer Hood, who credits Do Browser. It is published on GitHub under the MIT licence.
- Stars
- ★ 6.7k
- License
- MIT
- Cost
- Free
- Platforms
- Mac · Win · Linux
- Level
- Intermediate
Overview
Dev Browser is a command-line tool that lets an AI coding agent control Chrome with short scripts. The browser stays open between steps, so the agent can open a page once, read its structure, click or fill in forms, take screenshots and check the result. It can start its own browser or attach to a Chrome you already have open, and it installs a matching skill for Claude Code or Codex. It suits people who build or check websites with a coding agent.
Named pages keep their state between commands, so the agent works through a site step by step instead of starting over each time.
Try these first
Start with the first one: it is a small task you can check easily.
- Use dev-browser in headless mode to open example.com and tell me the page title.First try · read-only
- Take a screenshot of [page] and describe the layout.
- Open localhost:3000/signup, fill in the form with made-up data and report any error message.submits a form
- Connect to my open Chrome and list the tabs I have open.reads your browser
Example prompts written by Agent Flows Hub.
Features
From the creator's documentation.
| Feature | What it does | Runs |
|---|---|---|
| Persistent pages | Named tabs keep their state between scripts | Your computer |
| Page snapshots | A readable outline of the page for the agent to work from | Your computer |
| Screenshots | Saves screenshots of the pages it opens | Your computer |
| Interaction | Clicks, types and fills in forms on the page | Websites you open |
| Attach to Chrome | Connects to a Chrome that is already running, so scripts act in your own browser | Your browser |
| Agent skill | Installs a skill file so Claude Code or Codex knows how to use it | Your computer |
Good to know
You want a coding agent to check a web page the way a person would. The agent runs short scripts that open the page, read its outline, click or type, and report what changed, with the page kept open between steps.
Limits
- npm currently installs the stable 0.2 release, which runs on macOS, Windows and Linux. Version 1.0, in release candidates, supports only macOS and glibc Linux and changes the script helpers.
- Attached to your own Chrome, scripts can click, submit forms and act in the pages you have open. Start with a public page in a headless browser.
Notes by Agent Flows Hub (original project name: dev-browser), written from the creator's documentation, not from running it ourselves. Dev Browser: README · Dev Browser 0.2.9 release README
Stuck setting it up?
Tell us what happened. We take a look first; if it needs paid work, you get an estimate before anything starts.
More to write code
Other cli tools
- Fabric Runs ready-made AI prompts on any text, such as summarising an article into key points
- Horizon Turns the news sources you follow into a ranked daily briefing in English or Chinese
- LLM Lets you ask many AI models questions from the terminal and keeps a log of every answer
- MarkItDown Converts PDFs, Word, Excel, PowerPoint and more into Markdown text an AI can read
- OfficeCLI Lets your AI agent create and edit Word, Excel and PowerPoint files without Office
- OpenCLI Lets your AI agent use websites through the Chrome browser you are already signed in to