Say what you want in one sentence. Pip creates the project, breaks it into a handful of requests, and gives every one a file or folder to land in.
An agent works inside a real folder on your machine: writing files, running the checks, and fixing what fails. You see each step as it happens.
A tester has to pass before the work is done. Then it’s committed to real git history, so the next request starts from something that exists.
Pip’s memory is a locally stored vector database, not a chat log. Notes are embedded on your machine and kept as markdown files, and the index reindexes itself automatically whenever they change — so recall stays current and private.
