Docs

Wire up in five steps.

One gateway, one key format, OpenAI-compatible all the way down. Funders supply models, spenders call them at 66% of direct price.

Quickstart

01

Sign up

One account funds and spends. Confirm the email, open the dashboard.

02

Top up

Stripe top-up on the Wallet tab. Usage debits this balance at settlement — no balance, no requests.

03

Create a spender key

Spend tab → New key. Pick an alias, the models it may call, and a dollar budget with a window. The key is shown once — copy it now.

04

Point at the gateway

Swap the base URL to the one below, keep your code. Model names are the same canonical names funders supply.

05

Watch it meter

Usage tab shows requests, tokens, and the 66/75/25 split live. Settlement accrues every ~15 minutes.

SDKs

Same clients, new base URL. Gateway root: https://api.unburn.quantoxtinc.com

Python
from openai import OpenAI

client = OpenAI(api_key="sk-your-unburn-key", base_url="https://api.unburn.quantoxtinc.com/v1")
resp = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "hello"}],
)
Node
import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "sk-your-unburn-key",
  baseURL: "https://api.unburn.quantoxtinc.com/v1",
});
const resp = await client.chat.completions.create({
  model: "gpt-4o-mini",
  messages: [{ role: "user", content: "hello" }],
});
cURL
curl https://api.unburn.quantoxtinc.com/v1/chat/completions \
  -H "Authorization: Bearer sk-your-unburn-key" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-4o-mini","messages":[{"role":"user","content":"hello"}]}'

Coding CLIs

CLIs burn tokens fast — set a tight per-key budget first, then point them at the gateway like any OpenAI-compatible provider.

Codex CLIexport UNBURN_API_KEY=sk-your-unburn-key first
# ~/.codex/config.toml
model = "gpt-4o-mini"
model_provider = "unburn"

[model_providers.unburn]
name = "Unburn"
base_url = "https://api.unburn.quantoxtinc.com/v1"
env_key = "UNBURN_API_KEY"
wire_api = "chat"
OpenCodethen: opencode run --model unburn/gpt-4o-mini
// opencode.json
{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "unburn": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Unburn",
      "options": { "baseURL": "https://api.unburn.quantoxtinc.com/v1" },
      "models": { "gpt-4o-mini": { "name": "gpt-4o-mini" } }
    }
  }
}
Aiderper-key budgets stop runaway sessions
export OPENAI_API_BASE="https://api.unburn.quantoxtinc.com/v1"
export OPENAI_API_KEY="sk-your-unburn-key"

aider --model gpt-4o-mini
Claude CodeAnthropic-native — model must be a full provider/model string from live supply
export ANTHROPIC_BASE_URL="https://api.unburn.quantoxtinc.com"
export ANTHROPIC_AUTH_TOKEN="sk-your-unburn-key"

claude --model deepseek/deepseek-v4-flash-vision-exp
Claude Code quirk: it sends bare model names, but the gateway's Anthropic endpoint only routes full provider/model strings — always override with --model as above, using a name from live supply.

Billing, plainly

66

Spenders pay 66% of direct

Every request is priced off the provider's direct cost, metered by tokens. Top up with Stripe; settlement debits the wallet about every 15 minutes and itemizes it in the ledger.

75

Funders earn 75% of that

Earnings accrue into a withdrawable balance automatically. Withdraw $1.00+ to your bank via Stripe Connect after one-time onboarding.

25

Platform keeps the rest

The 25% covers rails, metering, and keeping the lights on. No subscriptions, no seats.

Funders

01

Add a key, we probe it

Paste a provider key, list the models it can serve. Each model gets a 1-token ping before going live — dead names never register. Keys are AES-256 encrypted at rest and never leave the server.

02

Cap what you sell

Set a total lifetime sell cap per key. Selling pauses automatically at the cap; raise it to resume with a fresh baseline. Rotate the secret any time without losing history or caps.

Troubleshooting

401 / invalid key

Wrong key or wrong base URL. Keys are shown once at creation — revoke and mint a fresh one if lost.

Key is blocked

The per-key budget bit. Raise the budget on the key or wait out the window.

Model not served

No funder currently supplies that model. Check the Spend tab supply list and pick a live one.

Requests stop mid-project

Wallet empty (top up) or the serving funder key hit its sell cap and paused — try another supplied model.