FAQ

Is Ageniti an agent framework?

No. Ageniti does not build agents, planning loops, memory systems, or workflow orchestration engines.

Ageniti helps developers build apps that agents can use.

Does Ageniti turn React components into tools automatically?

No. You explicitly define selected app capabilities as actions.

That is intentional. It keeps the contract reviewable and avoids accidentally exposing UI-only or internal behavior.

Can I use Ageniti without React?

Yes. The core action model and runtime are headless and work in plain Node.js or TypeScript projects.

React, Next.js, and Expo are adoption paths, not runtime requirements.

What should an action contain?

An action should contain a narrow app capability, typed input, typed output, risk metadata, documentation, and a run() function.

It should not contain routing, layout, animation, page state, or prompt orchestration.

What does an agent receive?

Depending on the surface, an agent or host receives structured schemas, metadata, and runtime results. Runtime invocation returns a stable envelope with ok, data or error, logs, artifacts, and meta.

Where should auth live?

In your app. Ageniti provides permissions, auth, user, and permissionChecker hooks so your existing authorization model can be applied consistently across surfaces.

What should I publish?

Publish the SDK package when you are maintaining Ageniti itself. Publish a generated app bundle when you are shipping one app's CLI/MCP/package artifacts for others to install.

Does the guide export call an LLM?

No. GUIDE.md is deterministic. It is generated from the app and action metadata you wrote.