qa-use exposes an MCP server that enables AI assistants (like Claude, VS Code Copilot, or Cursor) to drive browser automation and E2E testing through a standardized MCP interface. The server provides full browser automation capabilities, HTTP transport mode for web integrations, and a tunnel mode for backend-initiated testing. By starting the MCP server, you enable clients to request browser sessions, control navigation, interact with page elements, and run test sequences via the MCP protocol. The included tools cover client configuration for popular assistants, a web-friendly HTTP/SSE transport, and a robust set of browser automation commands that can be exercised from your AI agent or integration layer.

Prerequisites:

• Node.js (recommended: latest LTS) or a compatible runtime that supports npm/yarn

Installation steps:

1. Install the qa-use CLI globally (or use npx directly): npm install -g @desplega.ai/qa-use

   or run via npx without global install

2. Start the MCP server (default stdio) using the CLI: qa-use mcp

   Optional HTTP transport server:

   qa-use mcp --http

   Optional persistent tunnel mode:

   qa-use mcp tunnel

Notes:

• This MCP server is designed to integrate with AI assistants and browser automation workflows. If you prefer using npx directly without a global install, you can invoke the same command as shown in the Quick Start of the README.

Tips and common issues:

• Ensure your API keys and region settings are configured if you plan to use cloud or OpenAPI features.
• When using the HTTP transport mode, verify your network firewall allows WebSocket/SSE connections used by the MCP protocol.
• If the MCP server cannot reach the AI assistant, check environment variables and network proxies that might block outbound connections.
• The MCP server supports multiple clients; ensure you provide unique client IDs when configuring different assistants.
• Refer to MCP.md in the repository for a full reference of client configurations, transport modes, and the complete tools list.