Skip to main content
Neural Draft is the backend your AI-built site needs to be a real business. You build the frontend with Lovable, Claude Code, v0, Bolt, Cursor — whichever codegen tool you love. Neural Draft handles everything else: editable content, multi-language translations, the blog, social publishing, bookings, and a fully-wired Stripe Connect storefront. One project key. Five primitives. One SDK. One MCP server that teaches your AI tool how to use them correctly by default.

​
Live on Neural Draft

Real businesses currently running on Neural Draft:
  • LittleMind — AI-powered child mood tracking that turns 30 seconds of daily observation into emotional-pattern insights for parents during divorce, school changes, and other transitions.
  • Lolliland Rođendaonica — Children’s indoor playground and birthday-party venue in Rugvica, Croatia.
  • Lex Custis — EU AI Act compliance platform for law firms, with tamper-evident audit trails and regulator-ready reports for every AI interaction.
  • DevCred — Human-vetted developer credentials: a publicly shareable, manually reviewed ID that proves a developer is real.
  • Calma — Mindfulness platform with simple meditation and breathing practices designed for busy professionals.
You can use the same API + MCP server they use — see Quickstart.

Quickstart

From npm install to your first blog post in under 5 minutes.

MCP server

Drop @neuraldraft/mcp into Claude Code, Cursor, or Continue and your AI starts generating Neural Draft-native code automatically.

API reference

The full v1 surface. Try every endpoint inline with your live key.

Pricing

Pure usage-based credits. Hard stop at zero. No surprise bills.

​
The five pillars

Every Neural Draft project ships with all five — but you only pay for the credits you use. Mix one or all.

CMS

Translation keys, image keys, brand context, and a register-component flow that turns AI-generated HTML into editable sections without manual wiring.

Blog

Full posts, categories, tags, scheduling, and a research → write → image → translate → SEO pipeline behind a single async job.

Social

AI-generated multi-platform posts, scheduled publishing to Facebook, Instagram, X, TikTok, LinkedIn, YouTube — with brand voice baked in.

Booking

Bookable services, weekly availability, slot lookup, public booking flow, Stripe Checkout for paid bookings, and an embeddable widget.

Commerce

Products, variants, categories, orders, Stripe Connect onboarding, Stripe Checkout sessions, and drop-in storefront widgets.

​
Hello, world

A live example: create a draft blog post, in English, with one cURL. 
import { NeuralDraftClient } from "@neuraldraft/sdk";

const nd = new NeuralDraftClient({
  apiKey: process.env.NEURALDRAFT_API_KEY!,
});

const post = await nd.blogPosts.create({
  title: "Hello, world",
  content: "<p>The first post.</p>",
  language_code: "en",
  status: "draft",
});

console.log(post.slug);
Response (201 Created): 
{
  "id": 4711,
  "slug": "hello-world",
  "status": "draft",
  "title": "Hello, world",
  "created_at": "2026-04-19T10:14:22Z"
}
That’s the whole shape: a Bearer key, a JSON payload, an integer id, a stable slug. Every primitive in Neural Draft works the same way.

​
What makes this different

Most APIs pretend to support AI codegen by publishing an OpenAPI spec. We ship a real Model Context Protocol server (@neuraldraft/mcp) with resources, tools and prompts, so your editor’s AI generates code that follows our conventions automatically. No reading the docs. No hallucinated endpoints.
Every section your AI generates can be registered with one tool call — register_component(html, intent) — and instantly becomes editable in your end-customer’s admin UI. No CMS schemas, no migrations, no glue code. The AI puts data-translate="hero.headline" on text and the rest just works.
CMS, blog, social, booking and Stripe Connect commerce all sit behind one API key, on one bill, with one brand object that flows through every output. Mix one or all five — only credit usage differs.
No seats. No per-tenant tax. No “Pro” gate. You pay for the AI ops you run plus a small line item for the things that genuinely cost money to keep around (storage, connected social accounts, Connect accounts, custom domains). Hard stop at zero credits — no surprise bills.

​
Where to next

Quickstart

Five steps. Working API call. ~5 minutes.

Auth

API keys, scopes, rotation, common errors.

Errors

The full error catalog and retry guidance.

MCP setup

Wire Claude Code, Cursor or Continue in 60 seconds.

Webhooks

Event topics, signature verification, retry policy.

Migration

Already a SaaS tenant? Here’s the API-first path.

​
Community

Real questions get fast answers. The maintainers and most early customers hang out on Discord — drop a #hello and ask anything.

Join the Discord

Realtime chat, build-in-public, weekly office hours.

SDK on GitHub

File issues, browse source, send PRs.
The API is currently in public beta. We follow SemVer on the v1 surface — additive changes are normal, breaking changes are versioned and announced 90 days in advance in the changelog.