Approved AI skills
Skills agents are allowed to build with
Your team sets the rules once. Your coding agent writes the backend with the Xano SDK. Nothing ships without review.
Your coding agent writes the backend in TypeScript with the Xano SDK, you run it on the Xano Engine on your machine, then you deploy it to Xano, where your workspace rules gate the release.
Claude Code, Codex or Cursor writes tables, APIs, realtime channels, tasks, MCP servers and AI agents as TypeScript objects in your repo, with the Xano SDK. The xano-backend skill and the AGENTS.md that init writes keep it on your rules.
npx @xano/sdk init my-app && cd my-app A single binary with Postgres built in. No account, no network, starts in seconds. The engine version is pinned in package.json, so every checkout runs the same one.
npm run xano:deploy The same code lands in a disposable ephemeral and runs its tests. Only a stored release that came up and answered can be promoted, and promotion fails, naming anything that did not arrive.
$npx xanosdk deploy --ephemeral --test What your team mandates before anyone builds. Inherited automatically by every workspace.
Skills agents are allowed to build with
What every build must pass
The code that’s permitted in
How access works, every time
Where things run and data lives
Standing instructions agents inherit
Policy here isn’t documentation. Middleware enforces it on every request, before input validation and before the response leaves, assigned at the workspace level so no endpoint can skip it. Sensitive fields: encrypted in and out, no matter who built the endpoint.
Connect any agent (Claude, Cursor, Codex). Your agent scaffolds against your workspace rules by default. Deploying creates an immutable, sandboxed release. IT inspects and promotes the exact artifact to production.
Connect your coding agentSame agent. Different software. See Claude Code on Xano →
Experiments on Xano start within your policy. The one that works goes to production on the path that IT already trusts.
“When one of these becomes real, I'm not rebuilding it. It ships from where it started.”
Your coding agent writes it in TypeScript with the Xano SDK, you run it on the Xano Engine on your machine, then you deploy it to Xano. Start with npx @xano/sdk init my-app. On Xano, a build lands in a disposable ephemeral, runs its tests, and only a reviewed release is promoted to production.
Claude Code, Codex, Cursor, Copilot, Gemini CLI, Amp and OpenCode. npx @xano/sdk agent-skill install adds the xano-backend skill to each coding agent on your machine, so it writes valid Xano code in TypeScript and follows the AGENTS.md that init writes. Connect your coding agent.
Workspace rules are what your team mandates once and every build inherits, whichever agent is driving: approved AI skills, the test suite, allowed dependencies, auth patterns and identity, environments and residency, and agent rules. Every workspace inherits them automatically. How AI code governance works on Xano.
In the request path. Middleware assigned at the workspace level runs on every request, before input validation and before the response leaves, so no endpoint can skip it, whoever or whatever built it. Sensitive fields are encrypted in and out, and the release gate checks the rules again before anything promotes.
No. The Xano Engine runs your whole backend on your machine with no account and no network, so local deploys need no sign-in. You need an account when you deploy to Xano’s cloud, where npx @xano/sdk login connects your terminal. Get started with the Xano SDK.
Anyone builds with any tool. Nothing ships without review.