Getting started
Introduction
Inferiq gives your coding tools one API key for every model, and fix agents that turn an error into a cause and a patch. You pay per call from your balance, and every charge comes with a receipt.
What Inferiq is
Coding tools call a model on every step, and each provider has its own account, key, limits and API shape. Inferiq puts one OpenAI-compatible endpoint in front of them. Your tool sends a normal chat request with an Inferiq key, and the gateway:
- checks the key, the models it may use, its rate limit and its daily spending limit,
- checks that your balance covers the worst-case cost of the request,
- picks the model: the exact one you named, or the first available model behind an alias like
inferiq/balanced, - calls the provider, and moves to the next model if it is rate-limited, overloaded or down,
- returns the answer in the OpenAI format, streaming and tool calls included,
- charges the actual token cost and writes a receipt.
Provider keys stay on Inferiq’s servers. See How it works for the full path of a request.
What it’s for
One key for your coding tools
Point Cline, OpenCode, Cursor or your own code at one endpoint. Use Qwen3.8-27B, hosted by Inferiq, without a provider account or a GPU of your own.
A free key for each tool
Keys cost nothing. Give each tool its own key, limited to the models you want it to spend on.
Stay in the session when an attempt fails
A failed attempt is retried once before anything reaches your tool, and the model gets enough time to wake up after a quiet spell.
Know what every call cost
Each request gets a receipt: the model that answered, tokens in and out, cost, and your balance before and after.
Pay per call, no subscription
Requests are paid from a prepaid balance in US dollars at the model’s token price. A key can carry its own daily limit.
Call a specialist when your agent is stuck
Fix agents take an error and return the cause and a patch, for a price quoted before they run.
Who it’s for
| If you… | Inferiq helps you… |
|---|---|
| Code with an AI tool every day | Use every model through one key, and see what each session cost. |
| Run your own agents or scripts | Call one OpenAI-compatible endpoint with fallback and a receipt per request. |
| Keep hitting rate limits mid-session | Fall through to the next model instead of stopping. |
| Want spend you can predict | Pay from a prepaid balance with per-key daily limits, never a surprise invoice. |
What works today
Inferiq is in early access
The gateway works today:/v1/chat/completions with streaming and tool calls, /v1/models, /v1/receipts, aliases with three reasoning levels, free keys limited to the models you pick, balance holds, per-key daily limits and monthly budgets, and the fix agents (Quick Fix and Build Doctor) under /v1/fix. Deposits (ETH or USDG on Robinhood Chain, or ETH or USDC on Base) are live on mainnet. Claude Code connects through the Anthropic-format/v1/messages, and editors can call fix agents through the MCP server at /v1/mcp. Pages and sections marked with a phase describe what’s coming. See the roadmap.