Your AI writes the app.
Turn Zero is your backend.
1You ask
> Create a to-do app with sign-in and a database
in Claude Code or Codex
2Your AI calls Turn Zero Cloud
- ● create_application · todo-app
- ● read_library_entry database
- ● read_library_entry account
- ● deploy · version 1
✓ settled: deployed version 1
3It's live
todo-app-k3x9q2.ai.host
✓ hosted and monitored
Connect your MCP server from your computer. We'll email you the steps.
Free, Standard, and Pro plans to choose from. Paid plans free during our introductory period.
Connect the MCP server. It also creates your free account.
Claude Code
Codex
Claude desktop app
Type a prompt after connecting.
Every part of your backend, included
Your coding agent adds each part when your app needs it. It deploys to production, hosts and monitors your applications, and adds a development environment when you want one. The included feature packages ship test doubles, so your AI's tests run offline.
-
Node.js hosting
Your backend runs as ordinary Node.js code, WebSockets included.
-
Web front-end hosting
Your site's pages and files, served beside your backend.
-
Web, mobile, and desktop apps
One backend supports your web, mobile, and desktop apps.
-
PostgreSQL database
One for each environment, with migrations.
-
File storage
Named storage areas.
-
Sign-in for your users
The Turn Zero account service supports Google, GitHub, Apple, Microsoft Entra ID single sign-on (SSO), a passkey, or an emailed code, with open or invitation-only sign-ups.
-
Push notifications
Send to your users on iPhone and Android.
-
Email
Send confirmations, receipts, and invitations without running a mail provider. Pro plan.
-
Scheduled jobs
Recurring work on a schedule you declare, with every run recorded.
-
AI calls from your app
Use Gemini with your included AI allowance, or the provider you choose with your own API key.
-
Outside APIs on your keys
Call outside APIs on keys stored in a vault.
-
Logs and counters
Log messages and named counts that your AI reads back to find a fault.
-
Development and production
Try a change in development, promote it to production, and roll back if you need to.
-
Your own domain
Serve your production app on a domain you own. Pro plan.
-
Outbound firewall
Every outside call your app makes is checked against the hosts you declare.
An example session, shortened
- YouBuild a notes app where people sign in and save their notes.
- AgentBuilt. It is a Node.js server with a PostgreSQL database and sign-in, using the Database and Account feature packages. Its tests pass offline, before anything is deployed.
- YouDeploy this application to Turn Zero Cloud.
- AgentWhich plan should it start on: free, standard, or pro?
- YouFree.
- Agent
- create_applicationmyapp, on the free plan
- submit_manifesta database, and sign-in for its users
- store_secretan API key the app uses, typed into your own terminal and never into the chat
- deploystep: image_build at 0s…step: health_gate at 96s, shell_assets took 2ssettled: deployed version 1
- AgentVersion 1 is running in production at https://myapp-a1b2c3.ai.host. I requested its routes and read its logs: no errors.
Everything AI code needs to ship
Turn Zero closes the gap between AI-generated code and production-ready software.
Built for your coding agent
One MCP server gives Claude Code or Codex the whole backend: documentation it reads directly, ready procedures it can run, and feature packages with integration guides, and test doubles so its tests run offline.
Safe deploys
A new version takes over only after its health check passes, and a failed deploy leaves the running version serving. An earlier version is one rollback away. Add an optional development environment.
Keys stay out of your code
A secret is stored by name and never read back. An outside API's key is added to each call on its way out, so it never reaches your code, your chat, or your app's container. An outbound firewall checks every outside call against the hosts you declare: it records first, and refuses an undeclared host once Turn Zero switches your application to enforcement.
Your agent can watch it run
The agent that wrote the code reads its logs, its counters, and each step of a deploy, so it can find a fault and fix it. Risky actions wait for your approval in the browser. A CI job can deploy unattended with the turnzero command.