Documentation

Quickstart

Create a free account at /signup. Your $600 launch credit and 20M free inference tokens are applied automatically.
Open Dashboard → Security and create an API key. Store it securely; it is shown only once.
Call GET /api/v1/models to see the models available to your account.
Send your first request to POST /api/v1/chat/completions.
Watch usage and spend update in real time on the dashboard overview.

Authentication

All API requests use Bearer authentication. Include your key in the Authorization header as “Bearer YOUR_KEY”. Keys can be scoped, rotated and revoked at any time from Dashboard → Security. Never embed keys in client-side code.

Base URL

All endpoints live under /api/v1 on your platform domain. Requests and responses use JSON over HTTPS.

Use an OpenAI-compatible SDK

// Any OpenAI-compatible SDK works. Example in TypeScript:
import OpenAI from 'openai';

const client = new OpenAI({
  apiKey: process.env.AI_INFRA_API_KEY,
  baseURL: 'https://YOUR-DOMAIN/api/v1',
});

const res = await client.chat.completions.create({
  model: 'MODEL_ID',
  messages: [{ role: 'user', content: 'Hello, fabric.' }],
});
console.log(res.choices[0].message.content);

Core concepts

Projects: group resources, keys and budgets for a team or product.
Regions: choose from 13 regions to place compute and data where you need them.
Compute: CPU and GPU compute are prepaid. Choose an instance type, a count and the hours you need; paid time only counts while running, and unused time returns to your compute balance.
Metering and billing cycle: inference is metered per token and storage per GB-month, invoiced every 30 days.
VentureRails: report revenue events to route revenue and issue invoices for what you sell.

Next steps

Explore the API reference for endpoint details, browse our guides for step-by-step tutorials, or ask the AI ops copilot in your dashboard.