# MCP Browser

> MCP Browser is an open-source MCP server for browser automation. Browse login-protected and JavaScript-heavy pages through your own Chrome, Edge or Brave. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** CLI, vscode
- **Repository:** https://github.com/cherchyk/mcpbrowser/tree/HEAD/MCPBrowser
- **MCP Registry name:** io.github.cherchyk/mcpbrowser
- **Categories:** browser automation, web search and scraping
- **Listing updated:** 2026-11-15

## About

The idea behind MCPBrowser is that an agent should reach for it when plain HTTP fetches fail: pages behind SSO, CAPTCHAs or anti-bot checks, and single-page apps. It drives your real browser through Puppeteer, so you log in yourself once and the agent then extracts content from the authenticated session.

On top of raw browser control it adds an agent-oriented layer. HTML is cleaned to roughly 90 percent smaller by the README's account, React, Vue, Angular, Svelte, Next.js and Nuxt pages are detected for framework-aware waits, and login redirects are followed with a short SSO timeout before falling back to a longer manual one. Tabs are pooled by domain, and responses include next-step guidance on failure. A plugin system recognises known sites, with a Microsoft Word Online plugin documented.

The README warns that page content goes to your agent and then to its LLM provider, so only use it with parties you trust.

## Features

- **Real browser session:** Uses Chrome, Edge or Brave so you can log in normally.
- **Cleaner HTML:** Strips scripts, styles, SVGs and tracking attributes and absolutises URLs.
- **SPA detection:** Applies framework-aware waits for common front-end frameworks.
- **Auth handling:** Follows SSO and multi-step login redirects.
- **Form discovery:** browser_detect_forms lists fields and labels; browser_type_text fills several at once.
- **Domain tab pooling:** Reuses tabs by domain and re-reads the DOM without reloading.
- **Site plugins:** Known sites get specific actions, such as Word Online.

## Good for

- Read an internal dashboard behind corporate SSO
- Fetch a React app that needs JavaScript to render
- Log in and extract content after sign-in
- Fill a multi-field web form from an agent
- Run a page's JavaScript and capture a screenshot

## Install

### Claude Code (Terminal)

```sh
claude mcp add mcpbrowser -- npx -y mcpbrowser
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "mcpbrowser": {
      "command": "npx",
      "args": [
        "-y",
        "mcpbrowser"
      ]
    }
  }
}
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "mcpbrowser": {
      "command": "npx",
      "args": [
        "-y",
        "mcpbrowser"
      ]
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "mcpbrowser": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "mcpbrowser"
      ]
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "mcpbrowser": {
      "command": "npx",
      "args": [
        "-y",
        "mcpbrowser"
      ]
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "mcpbrowser": {
      "command": "npx",
      "args": [
        "-y",
        "mcpbrowser"
      ]
    }
  }
}
```

## Tools

- `browser_fetch_webpage`: Load a web page in the real browser, including login pages.
- `browser_execute_javascript`: Run JavaScript in the page.
- `browser_click_element`: Click an element on the page.
- `browser_type_text`: Type text into one or several form fields.
- `browser_get_current_html`: Re-extract the current page HTML without reloading.
- `browser_scroll_page`: Scroll the page.
- `browser_take_screenshot`: Capture a screenshot of the page.
- `browser_close_tab`: Close a browser tab.
- `browser_detect_forms`: Discover form fields, labels and constraints on the page.

## FAQ

### What do I need installed?

Chrome, Edge or Brave plus Node.js 18 or later.

### How can I install it?

As a VS Code extension, through Claude Code, Claude Desktop, or the mcpbrowser npm package.

### What licence is it under?

MIT, per the README badge.

### Is there a data risk?

Yes. Page content is passed to your AI agent and its LLM provider, and the README advises trusting both.

### Is MCP Browser free?

Yes. MCP Browser is open source, so you can use and self-host it for free.

### How do I add MCP Browser to Claude Code?

Run this command in your terminal: claude mcp add mcpbrowser -- npx -y mcpbrowser

### What are the best alternatives to MCP Browser?

The closest alternatives to MCP Browser are Browser Use, Scrapling MCP Server, Firecrawl MCP Server, SearXNG Search and Oya Browser.

## Alternatives

- [Browser Use](https://ai-powered.com/mcp/browser-use): Browser Use is a freemium MCP server for browser automation.
- [Scrapling MCP Server](https://ai-powered.com/mcp/scrapling-mcp-server): Scrapling MCP Server is an open-source MCP server for web search and scraping.
- [Firecrawl MCP Server](https://ai-powered.com/mcp/firecrawl-mcp-server): Firecrawl MCP Server is a freemium MCP server for web search and scraping.
- [SearXNG Search](https://ai-powered.com/mcp/searxng-search): SearXNG Search is an open-source MCP server for web search and scraping.
- [Oya Browser](https://ai-powered.com/mcp/oya-browser): Oya Browser is a freemium MCP server for browser automation.
- [Mantis](https://ai-powered.com/mcp/mantis-2): Mantis is an open-source MCP server for browser automation.
- [NeuralVerge](https://ai-powered.com/mcp/neuralverge): NeuralVerge is an MCP server for web search and scraping.
- [AI-Powered directory](https://ai-powered.com/mcp/ai-powered-directory): AI-Powered directory is a free MCP server for developer tools.

---
Source: https://ai-powered.com/mcp/mcp-browser-3
