Skip to main content
This release makes Architect more autonomous and more reliable — it can now test and fix your app on its own, deploy without a GitHub account, import an existing repo to continue building, and show you exactly where your credits go.

​
New

Added a testing agent that automatically tests and fixes the issues in your app. Turn on Test in the chat composer and the agent verifies your app in a real browser after each build, catches the runtime and console errors you’d otherwise run into, and fixes them on its own — so you no longer have to spot an issue and prompt Architect to fix it. The app you get back has already been checked. (Adds ~2–5 minutes per build; toggle it off anytime.)
Enabling the Test option in Architect's chat composer
Architect's testing agent verifying a build and fixing errors on its own
Deploy without connecting GitHub. You can now deploy immediately — no GitHub account required. Architect provisions and manages the repository for you using its own platform GitHub. Prefer to own the code? Use Export to my GitHub to move a platform-managed app into your own account whenever you like, and repositories are cleaned up automatically when an app is deleted.
Deploying an Architect app without connecting a personal GitHub account
Import a GitHub repo and keep building. Already have a Next.js project? Import any GitHub repository into Architect and continue building on it with Architect’s agents — no need to start from scratch. Architect brings your codebase into a fresh workspace so you can iterate, add agents, and deploy from there.
Selecting a GitHub repository to import into Architect
Configuring a GitHub repository import in Architect
Continuing to build on an imported GitHub repository in Architect
Rename your deployed app’s URL. Give your deployed app a custom, memorable subdomain instead of the auto-generated one. Pick a name that’s on-brand; Architect checks that it’s available and updates your app’s live URL.
Renaming a deployed app's URL in Architect
See exactly where your credits go. Usage now includes a detailed, per-agent breakdown of what each build consumed — across Plan, Agent Creator, UI Generation, Build, and Testing — alongside a token breakdown (input / output / cached) and your run and lifetime totals. The figures come straight from the sandbox’s credit ledger with your billing rate applied, so what you see matches what you’re charged.
Per-agent credit and token breakdown in Architect's usage panel

​
Improvements

More reliable previews & sandboxes. We’ve hardened the environment that runs your live preview and fixed the most common ways it used to break — previews stuck on “Connection lost,” 401 / “invalid sandbox” errors, and apps not serving on their preview port. An app is now marked ready only once its preview actually loads (no more false “ready”), a momentary control-plane hiccup is treated as a transient blip rather than a crash, and your app keeps running through stray restarts — so previews come up cleanly and stay up. Your PRD stays in sync on every iteration. The plan document now updates as you build. Each time you refine your app, Architect rewrites and persists the PRD to reflect the change — so the plan always mirrors the current state of your app, and every iteration carries accurate context. A more polished Architect, and better-looking apps. We’ve refined Architect’s overall look and feel, improved the visual quality of the apps it generates, and made the build chat more stable.