Cloudflare OS — Deploy & Usage Guide

A beginner-friendly, complete guide. Follow it and you'll have Cloudflare OS (Vercel edition) deployed to the public web, with all optional features configured.

What this is

Cloudflare OS is an AI coding workspace: describe an app in natural language and an AI agent builds it for you, with a live preview, file sharing, and external service connections.

⚠️ Derivative work: This project is developed based on (二次开发自) Cloudflare OS under the Apache License 2.0.

This version is a secondary development (二次开发) rebuilt for Vercel using Next.js + Postgres. It needs no paid Cloudflare plan and can be self-hosted for free.

Feature overview

Accounts & sign-in

  • Registration / login (password + GitHub + Google + email-code)
  • Cloudflare Access SSO gate (optional)

App building

  • Multi-file code editor (Monaco, same engine as VS Code)
  • AI Agent: build / modify apps with natural language
  • Live preview (iframe)
  • File history / version rollback (each change is recorded; restore with one click)
  • Realtime multi-user collaboration (Liveblocks: sync edits, peer count)

Share & collaborate

  • File sharing (Cloudflare R2, expiring links)
  • One-click static publish (build the workspace into a static site + public link, no external deploy)
  • Blueprint export / import (.gadget.json archive)
  • Public blueprint share links (viewable without logging in)
  • Favorite workspaces (star)
  • Workspace collaborators (invite by username, read-only / editable)
  • File-level shares (grant one file without opening the whole workspace)
  • Public context library (submit a doc → admin review → visible to everyone, agent references it automatically)
  • In-app + email notifications (collaborators, review results, ticket replies…)

Manage & extend

  • External connections (GitHub)
  • SSH persistent sessions (remembers the remote directory & environment across commands)
  • Scheduled tasks (cron expressions, auto-run with logs)
  • Context doc library (agent reference)
  • Admin panel (users / settings / AI / audit)
  • Audit log (exportable as CSV / JSON)
  • AI usage quotas (per-user / per-group daily limits, auto-throttled)
  • Multiple AI Providers (DeepSeek / OpenAI / local, etc.)

Documentation

Quick start (3 steps)

  1. Create a database: make a free Postgres on Neon and get DATABASE_URL
  2. Deploy: push to GitHub → import into Vercel, set Build Command to pnpm install && pnpm db:push && pnpm build
  3. Configure AI: Admin panel /admin → AI Providers → add DeepSeek, etc.

Full steps in the Deploy guide.