BrowserStack MCP Server logo

BrowserStack MCP Server

by MCP-Mirror

BrowserStack MCP Server enables developers and testers to use the BrowserStack Test Platform directly from their favorite AI tools. It allows for manual and automated testing of apps and websites on BrowserStack's real device and browser cloud.

View on GitHub

Last updated: N/A

What is BrowserStack MCP Server?

The BrowserStack MCP Server is a tool that integrates BrowserStack's Test Platform with AI tools, allowing users to perform manual and automated testing of web and mobile applications on BrowserStack's cloud infrastructure.

How to use BrowserStack MCP Server?

The server is installed via npm and configured with your BrowserStack username and access key. Configuration files are provided for VSCode, Cursor, and Claude Desktop. Once installed and configured, you can use natural language prompts within these IDEs to initiate testing actions on BrowserStack's platform.

Key features of BrowserStack MCP Server

  • Manual App Testing on Real Devices

  • Manual Web Testing on Real Browsers

  • Automated Testing (Playwright, Selenium, etc.)

  • Accessibility Testing

  • Test Observability

Use cases of BrowserStack MCP Server

  • Debugging app crashes on specific devices

  • Testing websites on different browsers and devices

  • Running and debugging automated test suites

  • Checking for accessibility issues

  • Porting test suites to BrowserStack infrastructure

FAQ from BrowserStack MCP Server

What Node.js version is required?

Node.js version >= 18.0 is required. Recommended version is v22.15.0 (LTS).

How do I get a BrowserStack account?

Sign up for a BrowserStack account on their website. Open-source projects may qualify for a free plan.

Where do I find my BrowserStack username and access key?

Your username and access key can be found in your BrowserStack Account Settings.

Which MCP clients are recommended?

Github Copilot or Cursor are recommended for automated testing and debugging. Claude Desktop is recommended for manual testing.

What if I encounter unexpected results?

Tool invocations rely on the MCP Client which in turn relies on an LLM, hence there can be some non-deterministic behaviour that can lead to unexpected results. If you have any suggestions or feedback, please open an issue to discuss.