create-lightnode-app

create-lightnode-app

Scaffold a new project with end-to-end encrypted LightChain AI inference in one command. Like create-next-app, for LightChain dApps.

npm create lightnode-app my-app

Pick a template, set a private key, you're inference-enabled in ~2 minutes.

First time? Read the Getting Started guide. It walks you through making a wallet, getting free testnet funds, and your first AI call with no prior experience.

Note: this command creates a new folder (my-app/). You must cd my-app before running npm install or npm run dev. Running them in the parent folder fails with npm error Missing script: "dev".

Want to add inference to a folder you already have instead of making a new project? Use npx lightnode-sdk add inference (see the SDK). That writes files into your current folder and does not create a subfolder.

Templates

Template	What you get	Run
node (default)	A 120-line index.ts that runs one end-to-end inference and prints the answer + three on-chain tx hashes.	npm start "your prompt"
nextjs-api	A Next.js 15 app with app/api/inference/route.ts wired up + a minimal page UI that calls it.	npm run dev
hono	A standalone Hono server exposing POST /inference. Deploys to Bun / Railway / Fly / any Node host.	npm start

Not sure which? Pick node to get to a first inference fastest (just a script + a key), nextjs-api if you want a web UI, or hono for a standalone API server you can host anywhere Node runs.

Non-interactive

npm create lightnode-app my-app -- --template nextjs-api --network testnet

Both --template and --network are validated; unknown values error out.

What gets installed

The scaffolded package.json already includes:

	Why
lightnode-sdk	The client (prepareSession, submitPrompt, decryptResponse, etc.)
viem	On-chain signing for createSession + submitJob
ws	Relay WebSocket (Node-side)

Plus framework deps for nextjs-api (next, react, react-dom) or hono (hono, @hono/node-server).

The project is SDK-ready the moment it scaffolds: lightnode-sdk (pinned to the current ^0.19.0 line; npm may lag the repo until the next publish), viem, and ws are already in package.json, so you just import { ... } from "lightnode-sdk" and start building. See the SDK README for the full API surface (read client, inference, the WorkerOperator, the CLI).

After scaffolding

my-app/
  .env.example       # set PRIVATE_KEY here
  README.md          # template-specific run instructions
  package.json
  ...

cd my-app
cp .env.example .env
# put a funded testnet (or mainnet) private key into .env
npm install
npm start            # or `npm run dev` for the nextjs-api template

Testnet inference is free - get LCAI at https://lightfaucet.ai. Mainnet inference costs ~0.022 LCAI per call.

Want to add inference to an existing project?

Use the CLI from the lightnode-sdk package instead:

npx lightnode-sdk add inference --template nextjs-api

It detects your stack and patches the right files. (Once lightnode-sdk is installed in the project, bare npx lightnode <cmd> works too; standalone, npx lightnode resolves to an unrelated npm package, so use npx lightnode-sdk.)

Where this fits

• Live playground: https://lightnode.app/playground
• Builder docs: https://lightnode.app/build
• SDK reference: https://www.npmjs.com/package/lightnode-sdk
• Source: https://github.com/marinom2/lightnode/tree/main/create-lightnode-app

License

MIT.