AI Integration Generator CLI

An interactive terminal wizard that generates a fully working, standalone AI integration project for any existing API — an MCP server, an AI chat server and a drop-in AiChat.jsx chat component — plus Docker configuration and documentation. All 8 steps run right in the terminal, with a modern create-vite / create-next-app style UI (gradient logo, diamond step tracker, guide rails, spinners).

Install

npm install -g ai-integration-generator-cli

Then run it anywhere:

ai-generate      # or the shorter alias: aig

How it works

The CLI is fully standalone — every part of the pipeline (connection probing, endpoint analysis, MCP tool mapping, project rendering, ZIP creation) runs inside the CLI process through a bundled generation engine. No backend API and no MongoDB are required — the only prerequisite is Node.js ≥ 18.

Everything is generated locally on your machine, in the terminal.

The 8 steps

| Step | What happens in the terminal | | ---- | ---------------------------- | | 1. Application | Project name, app URL, API base URL | | 2. Authentication | none / bearer / api-key / custom-header / basic (secrets are masked, never stored) | | 3. AI Configuration | Fixed provider/model (Groq + openai/gpt-oss-120b) | | 4. API Discovery | Point at a .txt file with your endpoints — routes are parsed automatically (GET /users, router.get('/x'), @app.post('/x'), bare /x lines) and shown in a table; manual add/remove and an optional live connection test are also available | | 5. Request Payloads | Per POST/PUT/PATCH endpoint: paste a JSON example and it is flattened into a payload table — nested objects are fully expanded (paymentInfo.id, paymentInfo.status, ...), with add / edit / remove / required / description per row | | 6. Preview | Shows tools, ports and the environment contract | | 7. Generate | Runs the whole generation locally — validating → analyzing → rendering → zipping, tracked live | | 8. Download | Copies the generated ZIP package(s) of your choice into a folder and can extract them locally |

Endpoints file

GET /users
POST /users
GET /users/:id
PUT /users/:id
DELETE /users/:id
POST /orders
router.get('/products', listProducts)

Relative paths are resolved against the current directory and the repository root, so ./app/routes.txt works no matter where you launch the CLI. ~ paths and quoted paths (Windows "Copy as path") are handled too.

Nested payloads

A pasted example like this fills the payload table with every child field:

{
  "product": ["6689163bb32037101ed659ed"],
  "user": "6a7d7e2939cf8829be3874ef",
  "quantity": 2,
  "paymentInfo": {
    "id": "payment_id",
    "status": "payment_status",
    "itemsPrice": "1899.00",
    "totalPrice": "1949.00"
  }
}

becomes rows product, user, quantity, paymentInfo.id, paymentInfo.status, paymentInfo.itemsPrice, paymentInfo.totalPrice — and the generated MCP tool schema keeps the same nested shape.

Options

-v, --version   print the version
-h, --help      show help

Environment variables:

• ALLOWED_PRIVATE_HOSTS — extra private hosts allowed for connection probes (comma-separated). localhost and loopback addresses are already allowed since the CLI runs on your own machine, e.g.:

  $env:ALLOWED_PRIVATE_HOSTS = "192.168.1.50"; ai-generate

What you get

<project>/
├── ai-server/         Express SSE chat API: Groq ──> MCP tools
├── mcp-server/        Streamable HTTP MCP server (tools + api://info resource)
├── ai-chat/           AiChat.jsx drop-in component + INSTALL.md
├── README.md          project documentation
├── docker-compose.yml both servers containerized
└── .env.example       one per service (secrets always empty)

See the repository at https://github.com/08Uzair/Ai-Integration-Generator for the backend and the full project.