Repolode
Submit
vercel-labs

vercel-labs/agent-browser

Native Rust CLI that lets AI agents drive a Chrome browser through snapshots, element refs, and common automation commands.

View on GitHubWebsite
Stars
44,312
Forks
3,028
Open issues
721
Last commit
9 hours ago

What is it?

agent-browser is a command-line tool for automating a web browser, designed so that AI agents can control it. It launches Chrome (downloaded through the built-in install command, or an existing Chrome, Brave, Playwright, or Puppeteer install that it detects) and exposes actions such as opening URLs, clicking, filling forms, taking screenshots, and reading page text. Its main interaction model is an accessibility-tree snapshot in which each element gets a short reference like @e2, so an agent can act on elements without writing complex selectors.

Beyond basic navigation, the CLI covers tab and frame management, network request interception and HAR recording, cookie and storage handling, saved auth state, batch execution of multiple commands in one invocation, and an axe-core accessibility audit. It can also run an MCP stdio server with selectable tool profiles, so MCP clients can use the same browser capabilities. React component inspection and Web Vitals measurement are available when the app is launched with the React DevTools hook.

Key features

  • 01Accessibility-tree snapshots with short element refs for acting on page elements
  • 02Chrome installation via Chrome for Testing, or detection of existing Chrome, Brave, Playwright, and Puppeteer installs
  • 03Commands for clicks, form filling, keyboard and mouse input, screenshots, PDF export, and video recording
  • 04Network route interception, request inspection, and HAR recording
  • 05Tab, frame, and dialog handling, plus saved auth state and cookie import
  • 06MCP stdio server with selectable tool profiles

Use cases

  • Letting an AI coding or research agent browse, fill forms, and extract content from websites
  • Scripting multi-step browser workflows in one batched CLI call
  • Running axe-core accessibility audits on a page or URL from the command line
  • Debugging React apps by inspecting component trees and render activity

Quick start

  1. 1

    Install the CLI globally via npm

    npm install -g agent-browser
  2. 2

    Download Chrome for Testing (first time only)

    agent-browser install
  3. 3

    Open a page, take a snapshot, and click an element by ref

    agent-browser open example.com

Things to know

  • Requires downloading Chrome for Testing before first use.
  • Linux system dependencies may need a separate install step that can fail.

Topics

#browser-automation#ai-agents#rust#cli#chrome#accessibility#mcp#playwright-alternative

FAQ

What is agent-browser?
Native Rust CLI that lets AI agents drive a Chrome browser through snapshots, element refs, and common automation commands. agent-browser is a command-line tool for automating a web browser, designed so that AI agents can control it. It launches Chrome (downloaded through the built-in install command, or an existing Chrome, Brave, Playwright, or Puppeteer install that it detects) and exposes actions such as opening URLs, clicking, filling forms, taking screenshots, and reading page text. Its main interaction model is an accessibility-tree snapshot in which each element gets a short reference like @e2, so an agent can act on elements without writing complex selectors.
Is agent-browser free to use? What license does it have?
agent-browser is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is agent-browser still maintained?
Actively maintained. The last commit to agent-browser was on Oct 20, 2026.
What can I use agent-browser for?
Letting an AI coding or research agent browse, fill forms, and extract content from websites Scripting multi-step browser workflows in one batched CLI call Running axe-core accessibility audits on a page or URL from the command line Debugging React apps by inspecting component trees and render activity
How do I install agent-browser?
Install the CLI globally via npm: npm install -g agent-browser Download Chrome for Testing (first time only): agent-browser install Open a page, take a snapshot, and click an element by ref: agent-browser open example.com
Where can I get agent-browser?
The source code and installation instructions are on GitHub: https://github.com/vercel-labs/agent-browser

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects