Log inSubmit server
Browser MCP logo

Browser MCP

Controls an existing Chrome session through MCP, with page interaction, screenshots, multi-session tabs, and human prompts.

InstallRepositoryWebsite
Full README Compare alternatives Embed widget Claim this listing

Install Browser MCP

Run the project’s own setup commands, in order.

Setup steps from the project

browser-mcp installs with its own commands rather than a config we can generate, so these are copied as-is. Run them in order in a terminal. Source: github.com

  1. 1. Install
    npx @agent360/browser-mcp install
  2. 2. Add to Claude Code
    claude mcp add --scope user browser-mcp -- npx @agent360/browser-mcp@latest
  3. 3. Configure
    2. **Toggle "Developer mode"** ON (top right corner)
  4. 4. Install
    3. **Click "Load unpacked"** (top left, next to "Pack extension")
  5. 5. Install
    1. [**Install the extension from the Chrome Web Store →**](https://chromewebstore.google.com/detail/agent360-browser-mcp/jdehgalffmffhfhmmhaokfbfnafnmgcl) - no Developer mode toggle, and Chrome keeps it updated for you.

Overview

browser-mcp connects MCP clients to the Chrome browser you already use, including its cookies, signed-in sessions, and open tabs. It exposes tools for navigation, page extraction, clicking, form entry, keyboard input, screenshots, JavaScript, file uploads, cookies, storage, and network waits. Human-in-the-loop prompts let an agent pause for CAPTCHA, 2FA, consent, or other decisions before continuing. Reach for it when browser automation must operate real authenticated websites, including sites that resist isolated Playwright or Puppeteer sessions.

Use cases

  • Automate authenticated web workflows in an existing Chrome profile
  • Read virtualized email, billing, search, or transaction lists
  • Test web applications through real clicks, forms, dialogs, and navigation
  • Pause for CAPTCHA, 2FA, consent, or other user-approved steps
  • Operate websites that do not provide a usable API

Key features

  • Controls real Chrome tabs through a local MCP server and extension
  • Supports page interaction with selectors, text targeting, and trusted mouse events
  • Extracts page content and long virtualized lists
  • Handles screenshots, JavaScript, cookies, storage, and console logs
  • Provides human prompts for CAPTCHA, 2FA, and sensitive decisions
  • Supports multi-session, color-coded Chrome tab groups

Tools40

What Browser MCP lets your agent do. Expand a tool for its parameters.

Verified live~9.8k tokens of context

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked Today

Recent health check succeeded.

26/31 checks healthy over the last 90d

What it can do

From the 40 tools this server returned when we ran it.

Runs commands · 2Can delete or modify data · 4Writes files · 1Handles credentials · 6Makes network requests · 5Read-only tools · 34
Network at startup3 connections
Files written at startupNone
Install scriptsNone
npm provenanceSigned build
Package age201 days
Maintainers1

Observed 2026-10-20 in an isolated sandbox. A description of reach, not a security audit.

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 18d ago via OSV.dev

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars62Stargazers on the source repository.
Last commit1d agoMost recent push to the default branch.
Tools exposed40Callable tools this server registers over MCP.
Directory activity13 viewsConfig copies, upvotes, and views on AllMCPs.

How Browser MCP works

What browser-mcp does

browser-mcp links a local MCP server and a Chrome extension so an agent can operate an existing Chrome session. The browser retains its normal cookies and login state, which allows workflows to continue through authenticated dashboards, webmail, social platforms, and services without creating a separate automation profile.

The server is intended for interactive browser automation rather than API-only access. It can inspect pages, manipulate controls, monitor navigation and network activity, and return screenshots or page data. Agents can also ask the user to complete a step in the visible browser, such as entering a 2FA code, solving a CAPTCHA, granting consent, or making a sensitive choice.

How it works

Two components are required: the local MCP process that the client launches and the Agent360 Chrome extension that drives the browser. The extension may be installed from the Chrome Web Store or loaded as an unpacked extension. The MCP server is installed from the npm package and registered in the MCP client's configuration.

Sessions can use Chrome tabs organized into separate, color-coded groups, allowing multiple agents to work at once. The server can list tabs belonging to its session, adopt or work with the active tab, and keep browser interaction tied to the user's real Chrome environment.

User prompts are displayed in the browser window. They can include fields when the agent needs information, while page navigation can cause prompts to be redrawn or completed with a navigation status.

Setup and configuration

Install the Chrome extension first, either from the Chrome Web Store or by loading the extension directory in chrome://extensions with Developer mode enabled. Then register the local server with an MCP client. For Claude Code, the documented command is:

Terminal
claude mcp add --scope user browser-mcp -- npx @agent360/browser-mcp@latest

Restart the client after registering the server. A setup command can also place the unpacked extension at ~/.browser-mcp/extension/ and register the server with supported clients it detects. The extension alone is not sufficient; it needs the local MCP server to connect.

Tools and capabilities

  • Navigate tabs and verify whether a page loaded, including handling blocked unloads and browser dialogs.
  • Read page text or HTML, target a CSS selector, and extract rows from long or virtualized lists by scrolling.
  • Click, double-click, right-click, hover, scroll, fill fields, select options, set dates, choose autocomplete values, and press keyboard keys.
  • Capture visible screenshots, execute JavaScript expressions, inspect console messages, and wait for elements or network requests.
  • Manage tabs, cookies, local storage, file uploads, dialogs, overlays, and user prompts.
  • Send HTTPS requests from the extension background when page CORS or CSP would block a page-origin request; cookies are not sent automatically.
  • Recover a stalled debugger attachment and use coordinate-based clicks when normal element targeting cannot reach a control.

Limitations and notes

Screenshots require a visible, uncovered Chrome window. Background tabs cannot provide wheel-driven loading or rendering behavior needed by some virtualized lists. Page reads are capped by default at 30,000 characters, so large documents may need selector-based reading or list extraction.

File upload requires files inside the MCP server's working directory and may require enabling file access for the extension. Native operating-system context menus are not opened by the right-click tool; it targets menus rendered inside the page. Browser requests made with browser_fetch require headers for authentication because the browser's cookies are neither sent nor stored.

The server is released under the MIT license and runs locally. Windows unpacked-extension installation is identified as not yet verified in the project instructions.

Read the full README View source on GitHub

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about Browser MCP

browser-mcp is an MCP server that connects AI clients to an existing Chrome session through a local server and browser extension. Its main tools navigate pages, read content, click and fill controls, capture screenshots, execute JavaScript, manage tabs and storage, and ask the user to handle CAPTCHA or 2FA steps.

Install the Chrome extension, register the local npm server with your MCP client, and restart the client. Claude Code can use `claude mcp add --scope user browser-mcp -- npx @agent360/browser-mcp@latest`; the extension can come from the Chrome Web Store or be loaded unpacked.

Yes, browser-mcp works with Cursor through Cursor's MCP configuration. The project also names Claude Code, VS Code, Codex, Windsurf, and Cline as clients that can run the local server.

Yes, browser-mcp drives the Chrome session you already use, including its cookies and signed-in sessions. This lets an agent work with authenticated sites without setting up a separate browser profile or login flow.

Yes, browser-mcp can pause and ask the user to complete CAPTCHA or provide a 2FA code, then continue in the same browser session. It also auto-clicks the reCAPTCHA v2 checkbox, with a human fallback for other CAPTCHA steps.

Related MCP servers

All in Browser Automation All alternatives
Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewBrowser MCP AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/browser-mcp?style=directory)](https://allmcps.com/mcp/browser-mcp)
HTML Embed
<a href="https://allmcps.com/mcp/browser-mcp"><img src="https://allmcps.com/api/badge/browser-mcp?style=directory" alt="Browser MCP on AllMCPs" /></a>