Your runtime or GitMir Cloud Compute
Keep execution where you want it. Intelligence is the common layer underneath the work, not a forced replacement for your stack.
Unless you add GitMir Cloud Compute, GitMir does not run your code. It reads a repository, answers questions about it, hands accepted work to whoever is doing it, and records what came back. The running — your editor, your agent, your pipeline, your servers — stays exactly where it is today, and GitMir sits underneath as the place to look the product up. GitMir Cloud Compute is for a project whose accepted work you would rather have GitMir carry out and commit.
Keep building, testing and deploying the way you do now
Let a person or an agent take an accepted task
Do the work in your own environment, with your own tools
Report back what the product now does differently
Rebuild the model and compare it with what was asked for
GitMir Cloud Compute is a paid add-on per project: it runs accepted work in its own sandbox and commits it or opens a pull request, but it never deploys. When an agent closes a task, GitMir stores what that agent said it did — it does not verify the claim; the check is you rebuilding the model and comparing. The one piece that does run on your side is a connector you download and run yourself, so a repository that cannot leave your network is read where it sits; it is handed out by key, not from a self-serve button.
See it on a real product.
Three models are published in full and open without an account. Ask one of them the question this page is about.