Serve

One npm package, every JS/TS framework. Import the adapter for your stack, point it at your ai.json, done.

npm install @aiendpoint/serve

Install

npm install @aiendpoint/serve

All framework dependencies are optional peer deps. You only need the one you use.

Express

import express from 'express'
import { aiendpoint } from '@aiendpoint/serve/express'

const app = express()
app.use(aiendpoint({ spec: './ai.json' }))
app.listen(3000)

GET /ai now serves your spec with proper headers.

Fastify

import Fastify from 'fastify'
import { aiendpoint } from '@aiendpoint/serve/fastify'

const app = Fastify()
app.register(aiendpoint, { spec: './ai.json' })
app.listen({ port: 3000 })

Next.js (App Router)

Create app/ai/route.ts:

import { aiendpoint } from '@aiendpoint/serve/next'

export const GET = aiendpoint({ spec: './ai.json' })

Hono

import { Hono } from 'hono'
import { aiendpoint } from '@aiendpoint/serve/hono'

const app = new Hono()
app.use(aiendpoint({ spec: './ai.json' }))
export default app

Works with Cloudflare Workers, Deno, Bun, and Node.js.

NestJS

NestJS uses Express under the hood. Use the Express adapter in main.ts:

import { aiendpoint } from '@aiendpoint/serve/express'

const app = await NestFactory.create(AppModule)
app.use(aiendpoint({ spec: './ai.json' }))

Options

All adapters accept the same options:

Inline spec example:

app.use(aiendpoint({
  spec: {
    aiendpoint: '1.0',
    service: { name: 'My API', description: 'Does things' },
    capabilities: [{ id: 'do_thing', description: 'Does a thing', endpoint: '/api/thing', method: 'GET' }]
  }
}))

Custom adapter

For frameworks not listed above, use the shared utilities:

import { loadSpec, cacheHeader } from '@aiendpoint/serve/handler'

const spec = loadSpec('./ai.json')

// In your framework's route handler:
response.setHeader('Content-Type', 'application/json')
response.setHeader('Cache-Control', cacheHeader(3600))
response.send(JSON.stringify(spec))

Generate ai.json: Use the CLI to create your spec from OpenAPI or interactively.