Outqube

Outqube is an AI agent that tests web apps in a real browser like a QA tester, records every step, and writes a failure report for each test that fails. This connector brings those results into Claude.

With it, Claude can:

• List your projects and recent runs, filtered by status, branch, commit or pull request
• Read a failed test's report: what failed and where, expected versus actual results, steps to reproduce, console errors and failed network requests, with the failing step's screenshot
• Open any step to see what the test agent did and why, the page URL, and that step's console and network entries
• Read a test's goal, steps and expected results, to tell a broken app from an outdated test
• Start a run of a suite, a set of tests or a tag on an environment, and cancel a run

Starting a run returns within seconds with the run's id and link; Claude can then check its progress and summarise any failures. Starting and cancelling runs needs Read and write access, which only organisation admins can grant. Everyone else connects with Read.

You sign in with your Outqube account and choose the organisation, the projects Claude can see, and the access level. You can revoke access at any time under Settings → Connected apps.