Vibetest Use
An MCP server that spins up several browser agents to test a site for broken links, UI bugs and accessibility issues
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- Drives a real browser and navigates sites
- Sends page data to the Gemini model
Install
Manual install
claude mcp add vibetest /full/path/to/vibetest-use/.venv/bin/vibetest-mcp -e GOOGLE_API_KEY="your_api_key"First run: uv venv, uv pip install -e ., playwright install chromium.
Security check
How we reviewMedium riskWe rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
- Drives a real browser and navigates sites
- Sends page data to the Gemini model
This is third-party code. Review the repository files before installing.
What it does
Vibetest runs an MCP server that, on a single command, launches a group of Browser Use agents and sends them across a given site. The agents look for UI defects, broken links, accessibility problems and other technical issues, then collect the findings into a report. It works on both a live site and local development on localhost. The request sets the URL, the number of agents and the window mode, headless or visible. The server connects to Claude Code and Cursor.
Who it is for. For developers and QA engineers who want a quick agent-driven pass over a site.
Good fit when
- You want a quick site pass for broken links and UI defects
- You want to test a local build on localhost before publishing
- You want to parallelize the check across several agents
Not a fit when
- You need a deterministic test suite with fixed scenarios
- You have no Google API key or access to the Gemini model
Example request
Run vibetest on localhost:3000 with five agents and collect a reportMCP
- Transport
- stdio
- Authentication
- API key
Environment variables
GOOGLE_API_KEYrequired, secret- Google API key for the gemini-2.0-flash model
Limitations
It needs a Google API key and runs on the gemini-2.0-flash model. It requires Python 3.11 and a Chromium install via playwright. Installation is from source with uv; there is no published registry package. Agent checks are non-deterministic, so results vary between runs.
How to disable. Remove the vibetest entry from the agent's MCP configuration and delete the project's local environment.
FAQ
Can it test a local site?
Yes, the server works with both a public URL and localhost, for example localhost:3000 or localhost:8080.
Why multiple agents?
The user sets the agent count, default three; more agents give a broader pass in a single run.
Related
MCP serversA CLI for testing and red teaming LLM applications: compares models, runs automated CI checks and finds vulnerabilities
Playwright CLI
playwright-cli
Microsoft's official Playwright CLI with an agent skill: drive a browser through short commands without heavy MCP schemas
The official MCP server debugger: web UI, automation CLI and terminal UI in one package
SonarSource's official MCP server: quality and security issues, quality gates and code analysis from SonarQube Server and Cloud
README in short
The README presents Vibetest as an MCP server for automated agent-driven QA of a site with Browser Use agents. Installation uses uv with an editable package install and a Chromium download for playwright. For Claude Code the server is added with claude mcp add and a Google API key, for Cursor through the MCP settings. The request sets the URL, agent count and headless mode. It needs Python 3.11 and a Google API key for the gemini-2.0-flash model. MIT licensed.