Two kinds of docs: guides for you (what you can ask your AI to
build) and the technical reference your AI reads over MCP. Every page is also
available as .md (markdown — best for LLMs) and .txt (plain text).
No code. What onvibe can do and how to ask your AI for it.
The big picture — you describe an app in plain words, your AI builds and deploys it, and you get a real link to share.
The one-time setup to connect Claude, ChatGPT, Cursor, LM Studio, or any MCP client to onvibe — with copy-paste config for each.
The kinds of apps onvibe is great at, each with a ready-to-copy prompt you can hand to your AI.
How to phrase your first request and how to iterate without breaking what already works.
Every app comes with a real database so it can remember things. What that means and what to ask for.
Two ways media works — files your app's visitors upload, and images you provide yourself (logos, photos).
How to give your app a Stripe key, an OpenAI token, or any other secret — without ever pasting it into the chat.
Have your app send transactional email — welcome notes, password resets, receipts — without setting anything up.
How to save restore points, roll back a change you don't like, and experiment safely with drafts.
Every app gets a real, public link you can share immediately. How it works and what to expect.
Have your app do something on a timer — daily reminders, weekly digests, nightly clean-ups — without you (or it) staying online.
Your dashboard, juggling multiple apps, access keys, and keeping things tidy.
Freeze a finished or live app so your AI can't change it by accident — and how to unlock it when you want to edit again.
What to do when your app shows an error or misbehaves — your AI can usually diagnose and fix it.
The practical boundaries — storage and file sizes, what onvibe is and isn't built for, and a few honest caveats.
The exact docs your assistant reads when building. Useful if you want to follow along.
Apps cannot be framed by default (anti-clickjacking). How to opt into embedding via security.frame in onvibe.json.
How to write main.ts correctly — mandatory export default pattern.
How to connect and use the PostgreSQL database from a deployed app.
How to evolve a PostgreSQL schema without breaking a running app: nullable/defaulted columns, keeping code and schema in sync, avoiding NOT NULL on populated tables.
End-user file uploads — private (uploadFile/getDownloadUrl, signed) or public (uploadFile with public:true, permanent cacheable URL) — vs developer static assets (upload_asset, public CDN). Read before building any upload feature.
How to capture browser-side JavaScript errors and view them via get_exceptions.
Error codes returned by tools and how to fix them.
How to deploy — always stage first, then deploy.
How to deploy a SvelteKit app (pre-rendered or SSR).
Clone your own project, or fork another user's project when they allow it.
How to run app code on a schedule (recurring tasks) — onvibe.json, create_cron, and the isCronRequest helper.