Turn Zero Cloud #

View this page as Markdown

Turn Zero Cloud is the complete production backend that hosts and runs applications built with AI coding agents such as Claude Code and Codex, with everything included on one predictable plan.

It connects to your AI coding tool through a Model Context Protocol (MCP) server. Once connected, your AI can code, build, and test backends, connect them to your app, and host your code in the cloud. Your AI provisions backend services, stores secrets by name, deploys your application, and reads its logs and stats directly. Turn Zero lets you keep coding with AI the way you do now. You can deploy straight to production, or turn on a development environment and promote from it, all from your chat session.

Where to start #

1. Connect Claude Code or Codex to the platform and sign in.
2. Use your AI coding tool to create and deploy apps. For example, type: "Create and deploy to turnzero a simple AI chat web app that lets me ask questions, with account sign in."

Features #

Built from the ground up for AI coding tools #

Deploy and run #

Backend feature packages #

Web app backends #

Mobile and desktop app backends #

The served version #

These pages describe one version of Turn Zero Cloud. Its number is in the page manifest, on the management contract's source page, and in every action's result as contract_version.

How it works #

Your AI coding tool writes the app; Turn Zero Cloud runs it, with hosting, database, sign-in and scaling included. Your tool writes, builds, and tests the application on your computer. The platform installs its runtime dependencies, builds the hosting image, runs it in production at the application's address, and records its logs, counters, and usage. A development environment is optional: once turned on, each deploy runs there first and a promote moves a version to production. Applications and environments explains the environments.

Three kinds of caller use the same actions: your AI coding tool through the MCP server, a script or pipeline through the HTTP API with a minted token, and you through the signed-in website. Each page shows the actions it uses in backticks, such as deploy and store_secret, as they appear in your tool. Each application has one plan; the pricing page gives prices and included capacity, and How the plans differ technically compares where each plan runs.

What the platform does not do #

The platform does not build your source, approve destructive actions for you, or make your application portable on its own.

Machine surfaces of this documentation #

This documentation is published at https://turnzero.ai/cloud/. An AI coding tool can read it without a browser through four machine surfaces:

• The index, at https://turnzero.ai/cloud/llms.txt, has one line per page with the page's title, address, and description.
• The whole text, at https://turnzero.ai/cloud/llms-full.txt, contains every page except the generated reference, which is read page by page.
• The page manifest is at https://turnzero.ai/cloud/docs.json.
• The Markdown copy of a page is at the page's address followed by index.md.

A fetch tool that summarizes what it reads, such as Claude Code's WebFetch, can leave pages out of its summary of the index, which lists every page. To find the page for a task, read Find the guide for a task, search the whole text, or, signed in, call read_documentation with a query.

At the origin root, https://turnzero.ai/llms.txt opens with what Turn Zero Cloud is and the site's public pages, then indexes this documentation tree, the pages under /cloud/. It also indexes the Turn Zero Docs tree, the pages under /tzdocs/. https://turnzero.ai/llms-full.txt contains the whole text of every documentation tree that every account can read. https://turnzero.ai/sitemap.xml lists every public page's address, the pages of this tree among them. Each hand-written page there carries the date its text last changed. The navigation beside each page lists pages alone, so these files do not appear in it, and their addresses do not change.

Every account can read this tree. The Turn Zero Team (tzTeam) tree is open only to an account with tzTeam access.

Turn Zero's production site is public, open to every visitor and to search engines. Its development site is private. Where a site is private, what a request receives depends on how it is sent:

• A browser opening a page, a request that sends Sec-Fetch-Mode: navigate or Sec-Fetch-Dest: document, receives the invitation notice. The response is 401 with a WWW-Authenticate header that mentions the request header Turnzero-Show-Page.
• The same browser request for a machine surface receives 404 with not found, never the notice.
• Any other request, such as a tool's fetch, a script, or curl, receives a documentation page or a machine file, marked X-Robots-Tag: noindex, nofollow. The site's own pages, such as / and /pricing, answer it 401 unless it sends Turnzero-Show-Page; the terms and the privacy policy answer it.
• A request that sends Turnzero-Show-Page with any non-empty value always receives the page or the file, and a Turn Zero Cloud bearer token always receives a machine file.

A tool signed in to Turn Zero Cloud reads the index as the resource docs://cloud/llms.txt and the whole text as docs://cloud/llms-full.txt. With the read_documentation action it also reads one page, one section, or the sections that match a few words, so it needs no download. Reading this documentation as a tool explains reading a long page in parts, going straight to a section, and searching.

Related #

The glossary defines each term this documentation uses, once. What happens without asking lists what your tool does on its own and which actions it asks you about first. The refusals page lists all of the platform's refusals, with the cause and remedy of each; a backend package's own failures are listed on its package page.

tzTeam is the companion plugin for keeping requirements, decision records, and raised items in files you own. Its documentation is a separate tree on this site, at /tzteam/, open to an account with tzTeam access.