What Agents Can Do
Connecting an AI assistant to CheckMate, and the limits it works within.
You can connect an AI assistant — Claude, or any tool that speaks the Model Context Protocol — to your CheckMate account. Once connected, it can read and organise your testing work by talking to CheckMate directly, instead of you describing everything to it and copying the answers back.
Useful things to ask for: “draft me twelve test cases for the new checkout flow”, “what failed in the last round?”, “set up next sprint's phase with the same testers as the last one” .
It acts as you
A connected agent has exactly the permissions you have, and no more. If you can't rename a phase because you aren't a manager of its project, neither can your agent. Everything it does is recorded against your name, the same as if you had clicked it yourself.
What it can never do
- Submit test results. Results are the record of a human actually testing something. An agent that could fill them in would make the record worthless.
- Manage your organisation. No inviting, removing, promoting or demoting people — those affect someone other than you.
- Change these settings. An agent cannot read or edit what it is allowed to do. That is decided in the app, by you.