# browser-use [Health: Active]

**Category:** 📂 Browser Automation  
**Repository:** https://github.com/browser-use/browser-use  
**GitHub Stars:** 119284  
**Views:** 7  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/browser-use

## Description
Control a real Chrome browser to complete any task: fill forms, extract data, book flights.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "browser-use": {
    "command": "uvx",
    "args": ["browser-use"]
  }
}
```

## Documentation

## What browser-use does

browser-use connects an AI agent to a browser so the agent can carry out tasks through ordinary web pages. Supported workflows include navigating sites, filling forms, extracting information, finding available appointments, selecting dates and times, and completing bookings. The repository describes an implementation covering 31 browser actions.

The browser can be running locally, hosted by Browser Use Cloud, or exposed by an existing remote browser through Chrome DevTools Protocol (CDP). This makes the server applicable to interactive desktop automation as well as Docker, CI, and serverless workflows when paired with a cloud browser.

## How it works

An agent supplies a natural-language task and uses browser actions to inspect and operate the page. Browser Use can run with a model selected by the developer, including OpenAI models through `ChatOpenAI` or Browser Use’s BU2 model through `ChatBrowserUse`. The Python example creates an `Agent`, assigns a task, runs it asynchronously, and reads the final result.

For Claude’s toolset integration, Browser Use acts as the browser driver. The browser may be local, a Browser Use Cloud browser, or a remote browser connected over CDP. Bash is available alongside the browser actions, allowing the integration to process data and write files on the SDK host.

## Setup and configuration

The open-source Python library requires Python 3.11 or newer. A typical project can install it with uv:

```bash
uv add browser-use
```

The sample setup loads environment variables from `.env` and uses an OpenAI API key with `ChatOpenAI`. Browser Use’s BU2 model and its cloud browser require `BROWSER_USE_API_KEY`; the same key can enable either optional service. Cloud-browser use is selected with `Browser(use_cloud=True)`.

Local execution opens a browser on the machine running the agent. Cloud execution removes the need to install Chrome and is documented for Docker, CI, and serverless environments. Browser Use Cloud is billed by browser hour and network traffic.

## Tools and capabilities

- Drive a real Chrome browser through AI-directed browser actions.
- Navigate pages and interact with web controls.
- Fill forms and complete multi-step browser workflows.
- Extract information from web pages.
- Connect to local, cloud, or remote browsers over CDP.
- Use Bash in the Claude toolset integration to process data and write files.

Browser Use Cloud adds stealth browsers, residential proxies, CAPTCHA solving, and a live-view URL. These are cloud-browser features rather than requirements for local browser operation.

## Limitations and notes

Browser automation depends on the target site and the browser environment. Cloud-browser usage requires a Browser Use API key and incurs usage charges. The Python library example uses an external model API key, while BU2 uses Browser Use’s own API key. Remote-browser operation requires a browser endpoint that can be reached through CDP.

The project presents three related paths: a hosted cloud agent, a CLI for automating browser tasks, and the open-source Python library. The MCP-facing Claude toolset is a separate integration that uses Browser Use as its browser driver.

_Full upstream README: https://allmcps.com/mcp/browser-use/readme_