Ready to zap those pesky website bugs with Vibetest-use MCP? This open-source gem, powered by Browser-Use, lets you unleash AI agents to crawl your site, catch 404s, dead buttons, and UI glitches in under a minute. I found it on GitHub, and it’s a total game-changer for automated testing. With just a single command, you can spin up multiple agents to stress-test your site—local or live. In this tutorial, I’ll show you how to set up Vibetest-use MCP, run tests on a local or live site, and make your QA (quality assurance) process a breeze. Let’s squash those bugs!
Why Vibetest-use MCP is a QA Superhero
Vibetest-use MCP is a Browser-Use-powered tool that automates QA testing by deploying AI agents to crawl websites and flag issues like broken links, unresponsive buttons, or UI errors. Built on the Model Context Protocol (MCP), it integrates with Browser-Use to control browsers (Chrome, Firefox, etc.) via Playwright, ensuring thorough testing. The GitHub repo highlights its ability to test any URL—live sites like https://example.com
or local setups like localhost:3000
—with configurable agent counts (e.g., 3, 5, or 10) for deeper scans. From testing a dev site to a production application, Vibetest-use MCP makes QA a walk in the park.
Setting Up Vibetest-use MCP
Let’s get Vibetest-use MCP running. This is beginner-friendly, and I’ll keep it clear!
1. Clone the Repository and Set Up Virtual Environment:
- Clone the repo and prepare a Python environment:
git clone https://github.com/browser-use/vibetest-use.git
cd vibetest-use
uv venv
source .venv/bin/activate # macOS/Linux
.venv\Scripts\activate # Windows
uv pip install -e .
- This installs dependencies and sets up the MCP server environment.
2. Configure Your Environment Variables:
- Vibetest-use MCP needs a Google API key for some features. Export it or add to
.env
:
export GOOGLE_API_KEY="your_google_api_key_here"
- Alternatively, create
.env
:
GOOGLE_API_KEY=your_google_api_key_here
- Get a key from console.cloud.google.com.
3. Run the MCP Server:
- Start the server:
.venv/bin/vibetest-mcp
- This launches the MCP server, ready to accept test commands.

I had this up in 10 minutes—quick and painless!
Example Use Cases for Vibetest-use MCP
Ready to test with Vibetest-use MCP? Here are three ways to catch bugs fast!
1. Testing a Live Website with Multiple Agents:
- Test a live site like
https://example.com
with 6 agents:
vibetest my website with 6 agents: https://example.com

- Five Browser-Code agents crawl concurrently, checking for broken links, UI bugs, or accessibility issues. Expect output like: “404 at /page-not-found, unresponsive button at /submit.”

2. Testing a Localhost Development Site:
- Test your local dev server (e.g.,
localhost:3000
):
run vibetest on localhost:3000
- This catches issues early, like a misaligned div I found on a test site!

3. Headless Testing with Custom Agent Count:
- For CI/CD, run headless with 10 agents:
run a headless vibetest on localhost:8080 with 10 agents
- Headless mode runs without visible browsers, logging results (e.g.,
--output results.json
). Perfect for automated pipelines.
These tests took <60 seconds each—Vibetest-use MCP is a speed demon!
Integrating Vibetest-use MCP with Claude Code Desktop
Want to run Vibetest-use MCP tests from Claude Code Desktop? Here’s how to integrate it.
1. Add Vibetest MCP Server to Claude Desktop:
- Use the Claude CLI:
claude mcp add vibetest /full/path/to/vibetop-use/.venv/bin/vibetest-mcp -e GOOGLE_API_KEY="your_google_api_key"
- This registers the server with Claude Desktop, passing your Google API key.
2. Verify Connection:
- Open Claude Desktop and type:
/mcp
- Check that the
vibetest
server shows as connected.
3. Run Vibetop Commands:
- In Claude Desktop, enter:
Vibetest my website with 5 agents: https://example.com
Run vibetop on localhost:3000
- Claude forwards commands to the MCP server, returning test results.
I tested a site via Claude Desktop, and it flagged a form bug instantly—super slick!
Using Vibetest-use MCP with Cursor
Prefer Cursor’s sleek interface? Here’s how to set up Vibetest-use MCP for testing.
1. Add MCP Server via Cursor Settings UI:
- Open Cursor Settings > MCP (left sidebar).
- Click Add Server or the “+” button.
- Add this config (adjust paths and key):
{
"mcpServers": {
"vibetest": {
"command": "/full/path/to/vibetop-use/.venv/bin/vibetest-mcp",
"env": {
"GOOGLE_API_KEY": "your_google_api_key"
}
}
}
}
- Save and restart Cursor if needed.
2. Use Vibetest Commands in Cursor:
- In Cursor’s chat or command interface, try:
Vibetest my website with 5 agents: https://example.com
Run vibetest on localhost:3000
Run a headless vibetest on localhost:8080 with 10 agents
- Cursor communicates with the MCP server, showing results inline.
I ran a headless test in Cursor, and the JSON output was crystal clear—QA heaven!
Customizing Vibetest-use MCP for Your Needs
Want to tweak Vibetest-use MCP? Here’s how to make it your own.
1. Adjust Agent Count:
- Use
--agents 2
for quick scans or--agents 10
for deep testing. More agents mean better coverage but higher memory use.
2. Target Specific Features:
- Edit
prompts.py
to test specific flows (e.g., checkout process):
prompt = "Vibetest with 3 the checkout flow on {website}. Verify cart updates and payment buttons."

- Run:
uv run vibetest --url https://shop.com --prompt "Test checkout flow"
3. Integrate with CI/CD (Ubuntu):
- Add to GitHub Actions:
name: QA with Vibetest-use
on: [push]
jobs:
test:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- run: |
curl -LsSf https://astral.sh/uv/install.sh | sh
uv venv
source .venv/bin/activate
uv pip install -e .
uv pip install playwright
uv run playwright install chromium
uv run vibetest --url http://localhost:3000 --agents 5 --output results.json
env:
GOOGLE_API_KEY: ${{ secrets.GOOGLE_API_KEY }}
4. Use DeepSeek for Smarts:
- Pair with DeepSeek:
uv run vibetest --url https://example.com --llm deepseek-chat
- Enhances reasoning for complex tests.
Exploring Vibetest-use MCP Features
Vibetest-use MCP is packed with QA goodies. Check these out:
- Parallel Testing: Run 10+ agents for <60s scans, as hyped on X.
2. Cross-Browser Testing: Test Firefox/Safari via playwright.config.ts
:
{ browserName: 'firefox' }
3. Deep Research Mode: Use mcp-browser-cli --deep-research
for UX analysis, saving Markdown reports.
4. Community: Join Browser-Code’s Discord or GitHub (38 stars) to share ideas.
I caught a Firefox-only glitch in a test—Vibetest-use MCP is a QA champ!

Wrapping Up: Master Vibetest-use MCP
You’ve unleashed Vibetest-use MCP to automate QA (quality assurance) like a pro! From setup to testing live sites, local servers, and integrating with Claude Desktop or Cursor, you’re set to keep your apps bug-free. Try experimenting with more agents, DeepSeek, or CI/CD, and be sure to document your APIs with APIdog. Share your Vibetest-use MCP wins on X or GitHub——I’m stoked to see your QA game soar! 🚀