A toolset designed for agents
With the Xano SDK’s typed surface, llms.txt grounding and the agent skill, your agents write against the real API instead of a guess.
The backend built for coding agents.
Your coding agent writes the backend in TypeScript with the Xano SDK, you run it on the Xano Engine on your machine, then deploy it to Xano. The SDK is a deterministic, fully typed surface that always compiles to a valid workspace, the agent skill sets up your coding agent to build on it, and Knowledge gives it the rules your business follows.
Claude Code, Codex, Cursor take turns driving the Xano Engine: reading tables and rows, seeding data, and running functions and tests through its local MCP server.
The guardrails your agents need
A Knowledge library tells your agents how to build the way your team builds.
The ability to see what agents are doing
Quickly verify what your agents built in one shared picture of the backend.
Three stops
Get started building with the Xano SDK and agents.
- 01 · Install
Install and connect.
Run
npx @xano/sdk agent-skill installto add thexano-backendskill to each coding agent on your machine: Claude Code, Codex, Cursor, Copilot, Gemini CLI, Amp or OpenCode. Already on Xano?npx @xano/sdk init my-app --from workspacereads your workspace back as TypeScript, ready to commit to Git. - 02 · Prompt
Prompt and build.
Open your project in your IDE with Claude Code running in the integrated terminal. Describe what you need (APIs, workflows, data models) and let it write the backend in TypeScript with the Xano SDK, grounded in the SDK’s
llms.txtand theAGENTS.mdthatinitdrops into every project. - 03 · Ship
Validate and ship.
Deploy with
npm run xano:deployto the Xano Engine on your machine, runnpx xanosdk test run-all, and open the builder link the deploy prints to review what your code compiled to. Then deploy it to Xano:npm run xano:deploy:ephemeralfor a live URL, and a release to reach your workspace.
Write with AI · Run locally · Deploy to Xano
Set up Xano in Claude Code, Cursor or Codex
Same SDK, same Engine, same deploy. Each guide covers what only that tool has.
Guide
Xano with Claude Code
Install the skill as a Claude Code plugin, connect to the Xano Engine’s MCP server, and follow the AGENTS.md that init wrote.
Xano with Cursor
A project rule that points Cursor’s agent at AGENTS.md, autocomplete across the SDK, and type errors in the editor before you deploy.
Xano with Codex
AGENTS.md from init, cloud tasks that run on the Xano Engine with no account or network, and headless sign-in for the cloud.
Any agent on the line
Work with your coding agent of choice.
What agents get
Context, guardrails, and a picture of what got built.
Agent grounding
A toolset designed for agents.
Turn your AI coding agent into a Xano expert. The Xano SDK is a deterministic, fully typed authoring surface: an agent, or you, emits well-typed TypeScript that always compiles to a valid, importable workspace.
- Ground agents in
llms.txtand its topic files, plus theAGENTS.mdthatinitdrops into every project. - Let agents read tables and rows, seed data, and run functions and tests through the Xano Engine’s local MCP server.
- Catch mistakes at compile time: a backend rename is a compile error, not a 404.
Knowledge
The business context your agents need.
With Xano’s Knowledge feature, you can give your agents all the information they need to build correctly and predictably.
- A workspace-native library of persistent, team-authored context tells agents how your team builds.
- Author Knowledge in the Xano UI, via the Xano Agent, or through external agents over the Developer MCP, with versioning, commit review, and admin governance built in.
- Your standards persist and carry across tools, so every agent builds the way your team builds instead of you re-explaining conventions each session.
- With the Xano SDK, knowledge() keeps that markdown as real .md files in your repo, reviewed in a diff and versioned with the code it describes.
Canvas View
Easily validate your agents’ work.
Your agents write TypeScript in your repo, and that code is the source. Deploy it, and Canvas View shows what it compiled to: a picture of what got built that is easy to validate.
- Validate workflows by enabling developers to check AI’s work without line-reading code.
- Easily identify logic errors before they run.
- Fix the TypeScript in
xano/and redeploy in seconds, with your agent or by hand.
“Every other tool makes you choose: move fast with AI, or build something you’d actually put in production. Xano is the first backend where I don’t have to pick.”
Docs
Related resources
Questions
Coding agents and the Xano SDK
What is the Xano SDK, and why use it with coding agents?
The Xano SDK is the official TypeScript SDK for Xano: the database, the APIs, even AI agents, written as typed TypeScript in your repo. It gives agents a deterministic, fully typed authoring surface, so what they write always compiles to a valid, importable workspace, and you review it in a pull request like any other code.
What is the agent skill?
xano-backend is an agent skill for the coding agent you already use: Claude Code, Codex, Cursor, Copilot, Gemini CLI, Amp or OpenCode. When a project needs a backend and has none, it has the agent confirm Xano with you, run init, deploy, and follow the AGENTS.md that init wrote. Install it with npx @xano/sdk agent-skill install.
What is Knowledge?
Knowledge is a workspace-native library of persistent, team-authored context that tells agents how your team builds. It holds a workspace AGENTS.md for always-on rules, docs for on-demand reference, and skills for step-by-step procedures. Changes go through commit review with full version history, and admins lock and enforce both editing and inclusion.
How do agents know the SDK’s API?
The package ships machine-readable grounding generated from the SDK’s own sources, so it cannot drift from them: llms.txt, the router with the mental model and the gotchas, plus the llms/*.md topic files it maps, and manifest.json with per-entry detail. init also drops an AGENTS.md brief into every project, so a coding agent writes against the real API instead of a guess.
Can agents work against a real backend?
Yes. The Xano Engine runs your backend on your machine with no account and no network, so an agent can iterate against a real backend without credentials. init connects Claude Code and Cursor to the engine’s MCP server, and npx xanosdk local mcp prints the server block for Codex and any other agent. There they read tables and rows, seed data, and run functions and tests. Code still changes in xano/ and ships with a deploy.
How do I check what an agent built before it ships?
Run the tests you author with npx xanosdk test run-all, or deploy and prove it in one step with deploy --test. The xano.lock every build writes keeps renames as renames, and the generated route manifest turns a backend rename into a compile error rather than a 404. In the builder, Canvas View shows what the code compiled to.
Do I have to write the code myself?
No. Your coding agent writes the backend in TypeScript with the Xano SDK, and you review it: in a pull request like any other code, and in Canvas View in the Xano editor, where your whole team sees what the code compiled to. You can still edit the TypeScript by hand whenever you want.
Can developers still code normally?
Yes. Developers write and edit the backend in TypeScript directly, use AI tools when helpful, and review the result in Canvas View, which shows what the code compiled to. The TypeScript in your repo stays the source.
I already have a Xano workspace. Can my agents work on it?
Yes. npx @xano/sdk init my-app --from workspace reads it back as readable TypeScript (real s.db.query(...) and f.email(), not a JSON dump), then checks its own work by diffing what it wrote against the workspace it came from. After a pull, xano/ is your source.
AI builds software. Xano makes it trustworthy.
The backend agents can build and humans can trust.