LetThemBuild for VS CodeClaude, ChatGPT, Gemini and Grok review each other's work in the checkout you already have open — and run the commands to prove what they claim. One of them leads. The others review what it did, and each reviewer runs its own commands rather than taking the lead's word: the build, the tests, whatever settles the question. Every verdict carries the commands it ran and whether they passed. What it does
Findings that name a real file and line become squiggles in the editor, attributed to the seat that raised them. Everything else stays in the panel — the markers are an addition to the transcript, never a replacement. RequirementsThe LetThemBuild desktop app, and its command line installed:
The extension runs that command. It does not reimplement the room, hold its own API keys, or talk to any model directly — the seats, keys, models and subscriptions are the ones you already configured in the app. A review started from the editor is the same review as one started from the app. If the command lives somewhere unusual, set Settings
What it deliberately does not doThe full transcript, the diff pane, the artifacts and the browser and simulator panes live in the app. This shows you the run and puts the findings where you are working; Open the full transcript in the app is one command away when you want the rest. Nothing is sent anywhere the app would not send it. |