Documentation

Quickstart

Create a key at /keys, export it, send a request. Both dialects accept the same key.

export FUELOFAI_API_KEY=foai_live_...

curl https://fuelofai.com/v1/messages \
  -H "x-api-key: $FUELOFAI_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-fable-5",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Explain a B-tree in two sentences."}]
  }'

import OpenAI from 'openai';

const client = new OpenAI({
  baseURL: 'https://fuelofai.com/v1',
  apiKey: process.env.FUELOFAI_API_KEY,
});

const res = await client.chat.completions.create({
  model: 'claude-fable-5',
  messages: [{ role: 'user', content: 'Explain a B-tree in two sentences.' }],
});

Authentication

Keys look like foai_live_…. Either header works, so an Anthropic-shaped and an OpenAI-shaped client can share one key.

Authorization: Bearer foai_live_...
x-api-key: foai_live_...

Endpoints

Anthropic-shaped tools usually want the root origin (https://fuelofai.com) and append /v1/messages themselves. OpenAI-shaped tools want https://fuelofai.com/v1.

Streaming

Set "stream": true. Frames arrive as Server-Sent Events in the format of whichever API you targeted, so existing parsers work unchanged.

curl -N https://fuelofai.com/v1/messages \
  -H "x-api-key: $FUELOFAI_API_KEY" \
  -H "content-type: application/json" \
  -d '{"model":"claude-fable-5","max_tokens":512,"stream":true,
       "messages":[{"role":"user","content":"Count to five."}]}'

Tool setup

Copy-paste configuration for the clients people actually use. Expand one for its exact snippet.

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://fuelofai.com",
    "ANTHROPIC_AUTH_TOKEN": "foai_live_..."
  }
}

Base URL:  https://fuelofai.com/v1
API key:   foai_live_...
Model:     claude-fable-5

model_provider = "fuelofai"
model = "claude-fable-5"

[model_providers.fuelofai]
name = "FuelOfAI"
base_url = "https://fuelofai.com/v1"
env_key = "FUELOFAI_API_KEY"

Base URL:  https://fuelofai.com/v1
API key:   foai_live_...
Model ID:  claude-fable-5

export OPENAI_API_BASE=https://fuelofai.com/v1
export OPENAI_API_KEY=foai_live_...
aider --model openai/claude-fable-5

models:
  - name: FuelOfAI
    provider: openai
    model: claude-fable-5
    apiBase: https://fuelofai.com/v1
    apiKey: foai_live_...

Base URL:  https://fuelofai.com/v1
API key:   foai_live_...
Model:     claude-fable-5

{
  "provider": {
    "fuelofai": {
      "npm": "@ai-sdk/openai-compatible",
      "options": {
        "baseURL": "https://fuelofai.com/v1",
        "apiKey": "foai_live_..."
      },
      "models": { "claude-fable-5": { "name": "claude-fable-5" } }
    }
  }
}

Base URL:  https://fuelofai.com/v1
API key:   foai_live_...
Model ID:  claude-fable-5

Base URL:  https://fuelofai.com/v1
API key:   foai_live_...
Model ID:  claude-fable-5

Response headers

Upstream-identifying headers are stripped. Your own dashboard logs show which channel served a request; the API response does not.

Errors

One envelope for every failure, whatever went wrong upstream.

{ "error": {
    "type": "insufficient_credits",
    "message": "Balance too low for this request.",
    "code": 402,
    "request_id": "req_8f2a..."
} }

Retry 429, 502, 503 and 504 with exponential backoff and jitter. Do not retry any other 4xx — fix the request instead.

Rate limits

Limits are enforced per key. Your current values are shown on your dashboard, and per-key overrides are available on request.

• A 429 always carries an accurate Retry-After. We never stall a request silently.
• Hard per-request caps: 10 MB body, 200 messages, 64 tools.
• max_tokens is clamped to the model's own output ceiling — see the model table.
• Free trial: 10 requests or 1,000,000 tokens, whichever runs out first.

Billing & privacy

Pay-as-you-go, metered per token at the published per-model rate. Cost is computed at full precision then rounded up to the nearest 0.000001 credit. Streaming places a temporary hold for the maximum possible cost and settles to the actual amount on completion; cancelled streams are billed only for tokens delivered. Full detail on the pricing page.