Donobu MCP: A Testing Agent for Your Coding Agent

Your LLM just shipped a feature.
Did it actually work?

In this video, we show how Donobu MCP turns AI-written code into AI-verified code — automatically.

Watch a founding engineer at Donobu walk through:

🧠 AI-driven, fully autonomous testing with Donobu Studio
🧪 Generating real Playwright tests from natural language
🎨 Verifying visual behavior (like dark mode, legibility, and color intent) — not brittle DOM selectors
🔌 Wiring Donobu MCP into Claude Code
⚡ Letting an LLM add a feature AND prove it works, end-to-end

We start simple:
“Search for fall leaves and verify the images are mostly red, yellow, and orange.”

Then we level up:

Testing a vibe-coded to-do app
1. Claude adds dark mode
2. Donobu MCP spins up a browser
3. Finds the toggle
4. Verifies the UI is actually dark
5. Confirms text is readable
6. Exports a self-healing Playwright test back into the repo

No screenshot diffing.
No fragile selectors.
No black-box SaaS.

Just: `page.ai.assert(...)`

This is what modern testing looks like when:

Your tests live in your repo
Your LLM writes code
And another AI verifies functionality

If you’re building with:

Playwright
MCP
Claude Code
LLMs in your dev loop

...and you want real confidence before shipping,
you’ll want to see this.

🔗 Download Donobu Studio: https://donobu.com/download
⭐ Star us on npm: https://www.npmjs.com/package/@donobu...
💬 Questions? Drop them in the comments

Because writing code is easy now.
Knowing it works is the hard part.