What Turn Zero Cloud is
Turn Zero Cloud hosts and runs the backend of an application your AI coding tool builds. Your tool reaches it through one Model Context Protocol (MCP) server, which covers backend services, hosting, deployment, and monitoring. The platform:
- provisions the services the application declares;
- keeps one platform credential per environment;
- applies network and execution limits; and
- records usage.
The services are a PostgreSQL database, file storage, end-user accounts, secret storage, and logging.
The lifecycle it covers
Your coding tool defines the application, writes its code, and builds and tests it on your machine. Turn Zero Blueprint can help maintain requirements beside that code.
Turn Zero Cloud accepts the built artifact, installs its runtime dependencies, builds the hosting image, and runs the application. The dependency install can run your dependencies' install scripts. A new application has one environment, production, and each deploy goes there. You can add a development environment at any time: a deploy then runs there first, and a promote moves a version from the history to production.
Your own machine can also run the application, against development data kept apart from production's; the glossary's local run entry explains how. You keep improving the application locally and deploy another artifact when it is ready. Deploy an application describes each step.
Who it is for
Turn Zero Cloud is for developers and AI-assisted builders who build with a coding tool and do not want to run servers themselves. Each application has one plan, chosen when it is created and changed with set_plan. The pricing page describes prices and included capacity.
One API, three callers
Three kinds of caller use the platform:
- an AI tool, through the MCP server;
- a script or pipeline, through the management HTTP API with a minted token; and
- a person, through the signed-in website.
All three use the same catalog of actions. What each caller may do depends on its credential, scope, and grants.
A destructive action needs a person's approval in the browser. A tool or token can request the action but cannot approve it; Management action tiers and approvals explains how. An action in the catalog that no build serves is left out of the tool list and answers 501 not_yet_provisioned, and the generated reference says which.
The vocabulary
The glossary defines the platform's terms, among them application, environment, manifest, custody, realm, pending action, and tier.
Where the platform stops
You supply the application's code and build tools. The platform does not approve destructive actions for you.
Your application uses PostgreSQL for its database and the platform's HTTPS interfaces for the other services. It needs no cloud provider's software development kit (SDK) for those operations. It still depends on the platform's service contracts and hosting limits, so using standard protocols does not by itself make it portable to another host.
Related
- Applications and environments explains the application record, its environments, and the version history.
- The manifest describes the declarations an application submits.
- Management action tiers and approvals explains the three tiers and the pending action a person approves in the browser.
- Sign-in, sessions, and tokens compares the credentials each of the three callers presents.
- Connect your tool is the first guide for an MCP client, after What your computer needs.
- Deploy an application covers the deploy, and the promote once a development environment is turned on.