vibetest-use

Vibetest Use is an MCP server that orchestrates multiple Browser-Use agents to automatically QA a website for UI bugs, broken links, accessibility issues, and other technical problems. It is designed to test both live websites and localhost development sites, providing a scalable way to run many browser-based checks in parallel. You can specify how many agents to run, the target URL, and whether to operate in headless mode. The server integrates with Browser-Use tooling to perform UI checks, navigation health, and accessibility verifications across pages, giving you rapid feedback on potential issues.

To use the server, first configure your mcpServers entry (as shown in the installation guide) and start vibetest. From there, you can train or run your tests by passing prompts or commands to the CLI, such as targeting a specific URL, setting the number of agents, and choosing headless vs. non-headless operation. The Quick Start examples show how to wire the server into Claude or Cursor, allowing you to manage and monitor the MCP from these interfaces. Results and statuses will reflect how many agents are connected and actively testing the provided site, as well as any detected UI or accessibility concerns.

Prerequisites:

• Python 3.11+ (as required by Vibetest Use)
• A working Python virtual environment (venv)
• Google Maps API key if you plan to use features requiring maps services (optional for Browser-Use integration)
• MCP-capable client interfaces (e.g., Claude or Cursor) for easier server management

Step-by-step installation:

1. Create and activate a Python virtual environment

   • python -m venv .venv
   • source .venv/bin/activate # on Unix/macOS
   • .venv\Scripts\activate # on Windows

2. Install the Vibetest Use package in editable mode (from repo path)

   • uv pip install -e .

3. Install the browser dependencies needed by Browser-Use (Chromium via Playwright)

   • uv pip install playwright # if not already installed via the package
   • playwright install chromium --with-deps --no-shell

4. Ensure you have a Google API key if required by your workflow and set it in the environment

   • export GOOGLE_API_KEY=your_api_key # Unix/macOS
   • set GOOGLE_API_KEY=your_api_key # Windows

5. Run or register the MCP server using your preferred orchestration (e.g., Claude or Cursor) as indicated in the README

   • Example registration (as shown): claude mcp add vibetest /full/path/to/vibetest-use/.venv/bin/vibetest-mcp -e GOOGLE_API_KEY="your_api_key"

Notes and tips:

• The server supports configuring the number of agents and target URLs via prompts or by editing the MCP config in Cursor. Common flags include specifying the URL and the number of agents for parallel testing.
• If you encounter connection issues, verify that the vibetest-mcp executable path is correct and that the virtual environment is active when launching the server.
• Ensure Playwright chromium is installed properly to avoid runtime errors when launching browser instances.
• If using Google API keys, keep them secure and do not commit them to version control.
• The npm_package field is null for this Python-based MCP server; there is no corresponding npm package to install.
• For debugging, check that the environment variable GOOGLE_API_KEY is accessible to the MCP process and that the path to the venv’s vibetest-mcp is correct in the mcp config.