Browser AI

The Browser AI is an MCP server for browser automation that lets you control web browsers using natural language through Claude.

It’s built for tasks like automated testing, web scraping, or just handling repetitive online chores without writing a single line of code.

Featuress

• 🗣️ Natural Language Control: Run browser automation tasks with plain English commands.
• 🤖 Claude Integration: Connects with Claude’s AI using your existing subscription through OAuth.
• 🗺️ Intelligent Tool Mapping: The AI figures out which Playwright tools to use based on your commands.
• 🧠 Session Management: Keeps track of what you’re doing across multiple commands.
• 🛠️ Smart Error Recovery: Provides clear feedback when something goes wrong.

Use Cases

• Rapid Prototyping and Testing: You can quickly test a new user signup flow by just telling the browser what to do: “Go to the registration page, fill out the form with test data, and submit it.” This avoids the need to write a full automation script for a simple check.
• Empowering Non-technical Team Members: A QA tester or product manager who doesn’t code can run browser-based tests. They can describe a user journey in English to verify functionality without needing a developer to write a script for them.
• Quick Data Scraping: If you need to pull some text from a website, you can direct the browser with commands like “Navigate to example.com, find the contact section, and copy the address.” It’s a faster alternative to writing a custom scraper for small jobs.
• Automating Repetitive Tasks: For any multi-step process you do frequently in a browser, like downloading daily reports, this server can handle it after you describe the steps once.

How to Use It

Installation

1. Install the server automatically using Smithery.

npx -y @smithery/cli install @qckfx/browser-ai --client claude

If you prefer a manual approach, you can install the package directly with npm:

npm install -g @qckfx/browser-ai

Or, you can run it directly without a permanent installation using npx:

npx @qckfx/browser-ai

2. The server needs to connect to Anthropic’s API. You have a couple of options for this.

OAuth Authentication (Recommended)

If you have a Claude subscription, this is the way to go. It uses your subscription credits for API access. Run the auth command:

npx @qckfx/browser-ai --auth

A browser window will pop up for you to log in and authorize the connection. The token is saved securely for future use.

API Key Authentication

As an alternative, you can use a standard Anthropic API key. You’ll need to set it as an environment variable:

export ANTHROPIC_API_KEY="your-api-key-here"

3. After authenticating, you need to let Claude know about the server.

For Claude Code:

Add the server with this command:

claude mcp add browser-ai -- npx --yes @qckfx/browser-ai@latest

If you’re using an API key, include it in the command:

claude mcp add browser-ai --env ANTHROPIC_API_KEY="your-api-key-here" -- npx --yes @qckfx/browser-ai@latest

For Claude Desktop:

You’ll need to edit your Claude Desktop configuration file (claude_desktop_config.json). Add the following JSON object:

{
  "mcpServers": {
    "browser-ai": {
      "command": "npx",
      "args": ["@qckfx/browser-ai"],
      "env": {
        "PLAYWRIGHT_MCP_PATH": "@playwright/mcp@latest"
      }
    }
  }
}

4. Once it’s all set up, you can start giving commands to Claude. Here are a few examples of what you can say:

• “Go to example.com and wait for the page to load”
• “Click the login button”
• “Fill the username field with ‘[email protected]'”
• “Take a screenshot of the current page”
• “Fill out the registration form with test data and submit it”

5. The AI interprets your command, converts it into the necessary Playwright tool calls, and the server executes them in the browser.

FAQs

Q: What is the main advantage of this over just writing a Playwright script?
A: Speed and accessibility. This tool is for getting quick automation tasks done without the overhead of writing, debugging, and maintaining a script. It also allows people who aren’t automation engineers to perform browser-based tests and tasks.

Q: Will the API usage be covered by my Claude subscription?
A: Yes, if you use the recommended OAuth authentication method, the server’s API calls will use your subscription credits. If you opt for an API key, usage will be billed to that key’s account instead.

Q: How does it handle complex, multi-step instructions?
A: The server maintains context during a session, so you can chain commands together. For instance, you can tell it to navigate to a site, then click a link, then fill a form in separate commands. For very intricate workflows, breaking the task into smaller, distinct commands generally gives the most reliable results.

Q: What happens if the AI misunderstands my command or an element isn’t found?
A: The server includes error handling and will return feedback, often directly from Playwright (e.g., “Error: selector not found”). For trickier issues, you can run the server with the --debug flag to see verbose logs of the specific tool calls the AI attempted to execute.