Overview of the BrowserStack MCP server

Integrate BrowserStack’s real device cloud with your AI agent to test, debug, and manage your workflow using natural language.

About the MCP server

The BrowserStack MCP server connects AI agents and Large Language Models (LLMs) to BrowserStack’s testing infrastructure and products from where you code, such as VS Code, Claude, or Cursor.

You can manage test cases, run manual and automated tests, debug issues, and get AI-suggested code fixes by writing natural language prompts. Because BrowserStack’s tools are available in your AI agent, you can test without switching between your IDE and BrowserStack.

Key features

The BrowserStack MCP server provides the following capabilities in your development environment:

• Natural language testing: Manage, execute, and debug your testing workflow using English prompts.

• Manual app and web testing: Launch your mobile apps or websites on thousands of real iOS and Android devices or desktop browsers for manual testing and debugging, including locally hosted projects.

• AI-powered automated testing: Run your Playwright and Selenium test suites on our infrastructure. Use AI-powered debugging to analyze failures, fetch logs, identify root causes, and apply fixes directly from your IDE.

• Integrated accessibility scans: Catch and fix accessibility issues early in your development cycle. Run scans on local or production websites to ensure WCAG and ADA compliance with AI-suggested fixes.

• Comprehensive test management: Create and organize projects, write test cases, trigger test runs, and update results within BrowserStack Test Management using natural language commands.

• Access to BrowserStack AI agents: Use agents to generate test cases from product requirements documents (PRDs), convert manual tests to low-code automation scripts, and automatically heal flaky tests.

How it works

The BrowserStack MCP server creates a secure, stateful connection between your AI model and the BrowserStack platform.

Here’s how it works:

1. AI client (your IDE): You issue a command in natural language to your AI agent, for example, “run my tests on an iPhone 15”.

2. BrowserStack MCP server (running locally): The MCP server is a process running on your machine. It receives the command, authenticates using your BrowserStack credentials, and translates it into a specific API request for the BrowserStack platform.

3. BrowserStack cloud: The server sends the request to the right BrowserStack product, such as App Live or Automate. Our cloud executes the test on a real device and sends back the results, logs, or session links.

4. Response to client: The MCP server forwards the response from BrowserStack back to your AI client, which presents it to you in a human-readable format.

This local server model keeps your BrowserStack credentials secure on your machine while providing a fast, interactive connection to the cloud.

The remote MCP server runs on BrowserStack’s cloud instead and uses OAuth to authenticate, so you don’t install anything locally.