We felt this was the natural next step for AI coding assistants. And with new asynchronous assistants like codex, having a plug & play tester makes sure you don't have to spend time debugging and reviewing AI's code and can keep up with the speed with which AI codes.

For basic QA testing, we have - UI, visual regression testing - save tests for regression - Self healing - persistent browser auth states across tests

For security analysis, we have - SAST - DAST

By the time we had already written 10k lines of code, I realised we should launch the product asap so we know what features people actually want.

All this makes sure that you just give the initial prompt, and then directly interact with the final working code. You can be just a bit more sure before deploying vibe coded SaaS since you'll have tested and secured all the basic stuff even if you don't know how to code.

We have tested this with the free version of gemini-2.0-flash and the results are good :) We use it daily in our workflows to build websites and as a final check before committing anything. It pinpoints errors, makes sure previous working features don't break and tells cursor if there is a common security issue and how to fix it.

We also limit token usage by smartly filtering and augmenting the DOM context from websites. This was achieved by modifying the browser-use code heavily to suit to a testing framework. Special thanks to them for their stellar open source work :]

Although you can use it locally and completely free from GitHub, a cloud based solution is also available on our website - (groundng.site) for $20. We appreciate any and all feedback to make this better.

Thank you all for your time :)

PS: We also launched on ProductHunt today [https://www.producthunt.com/posts/vibeshift-mcp].