VibeKit
Enter App
Code on iPhone with AI
VibeKit is a native iOS app for AI coding. You chat with a persistent agent — it writes the code, runs the build, and deploys to a live URL. Real Next.js, Express, and static apps. Real GitHub repos. No laptop required.
You don't type code. You direct the agent.
Coding on a phone keyboard is miserable — nobody disputes that. But code on iPhone with AI is different: you describe what you want in plain English, and the agent does the typing. Think of it like running a senior contractor: you point at a feature ("add a sign-in screen with Apple Sign-In"), they go do it, they come back with a working build. The phone is a control panel for a real engineer in the cloud, not a code editor with autocomplete.
Real apps, not toys
Every VibeKit app gets its own AWS Fargate container, EFS workspace, and nginx subdomain at <yourname>.vibekit.bot. The agent writes to a real Git repo on your GitHub (or our shared org if you don't have one), the container rebuilds on every push, and traffic is real-world HTTPS with auto-SSL. You're not playing in a sandbox.
What you can build from your phone
- Landing pages, marketing sites, blogs. Static or Next.js. Often live within 60 seconds of the first chat message.
- Internal dashboards, CRUD apps. Next.js + database connection from the agent's first turn.
- Backend APIs. Express / Fastify. Hook them up to your iOS app or a partner's frontend.
- Personal automations. Cron-style scheduled tasks the agent owns. Useful for scraping, email digests, light ETL.
- Imported repos. Connect an existing GitHub repo and the agent picks up where your laptop left off — useful when you're traveling and want to fix a bug from your phone.
How a session actually looks
yourname.vibekit.bot in Safari. Your app is live, on the public internet, in seconds.Why not just use ChatGPT on iPhone?
ChatGPT can write code, but it can't deploy. It can't see your app's runtime errors, can't restart your container when it crashes, can't pull from your GitHub or push back to it. VibeKit's agent has persistent context for each app (file tree, recent changes, build state) and can actually act on the environment. It's the difference between "here's some code, copy it into Xcode" and "here's a working deployment at this URL, want to keep iterating?"
What it costs
Free tier: 1 app, 256 MB RAM, 1.5 GB workspace, 10 AI sessions/mo. The free Qwen AI provider runs on OpenRouter's free pool (zero AI cost), so the free tier is genuinely free — not "free trial, then $20/mo." Bring your own Anthropic / OpenAI key for premium models with zero markup. Paid plans start at $19/mo (Builder, 3 apps + custom domain). Full pricing →