Thanks. Sorry for delay in response. The runner on the server runs using JS playwright. But from your local, if you have a python code, you can just connect directly to the mcp, with auth in the headers. I ll add that to docs, good point. Auth is basically just the username and the password.

Update since last week. Upon first feedback, have now added chat agent on the UI as well. The agent runs on a queue, which means, the tasks such as self write or heal tests, keep running on server, with max steps, max time set.

There was also a playwright feature that calculates coverage, which we have also added, in settings and viewer in built.

Will be looking forward to feedback on how we can further improve it.