AICoin

What's an AICoin?

1 AICoin always buys exactly 1 paid AI API call — a chat completion, an image generation, a text-to-speech clip — through proxy.aicoin.oeaio.com, our relay (aicoin-proxy) that holds its own paid provider credentials so you never need your own OpenAI/Anthropic/Google/etc. key. That 1-coin-per-call rate is fixed and enforced server-side, not a suggestion.

The "current market rate" above is a separate, informational number: a recency-weighted average of what a call has actually cost us in real provider spend recently. It's what we use to price coin packages fairly — it has no effect on the fixed 1-coin-per-call spend rate.

Getting AICoin

Receive a transfer

Anyone can send AICoin to your wallet address, from the wallet page or from within a supporting app. There's no real-money rail on this path — it's a direct peer-to-peer balance transfer, the same way you'd hand someone cash.

A small free-coin faucet also exists for trying things out (up to 10 coins per wallet per hour, while the shared pool lasts — … left) — it's for onboarding, not a way to fund real usage.

Ask for what you want, not for a provider

Three endpoints take a prompt and pick the model themselves: POST /text, POST /image, POST /audio. The relay tags what the request is about — code, legal, science, translation — and sends it to whichever provider is rated highest at that, skipping any it holds no key for or that failed its last liveness check.

curl https://proxy.aicoin.oeaio.com/text \
  -H "X-Api-Key: <token from /wallet>" \
  -d '{"prompt":"Why does this Java function throw a null pointer exception?"}'

{"answer":"...","subject":"code","provider":"openai",
 "considered":["openai","anthropic","kimi","google","mistral"],
 "calls":1,"coins_charged":1}

Every answer names who wrote it and who else was in the running: routing chosen for you is only worth having if you can see what was chosen. Images and speech come back in one shape whatever the provider answered in. The ratings are our opinion rather than a benchmark — name a provider in the request and no routing happens at all.

When one model isn't enough

A model answering /text on its own can decline to answer on its own. If the request genuinely needs several — a judgement that belongs to something that has read it, not to a rule about its length — the same request goes to the full panel below, merged and reviewed. It costs several times one call, so the model is told the price, and "escalate": false keeps any request to exactly one call.

see the whole ranking →

Supported AI providers — live connection health

The relay asks each provider directly, about once a minute, whether it is there. Live = it answered the last check. Degraded = it is answering but throttling us, or recent real calls came back rate limited or rejected for billing. Down = no answer, an error of its own, or the relay's key refused. Not configured = no key for this provider yet.

Checking…

full status page →

Integrate AICoin (for developers)

Building an iOS/macOS app? The shared AICoinKit Swift package (in the aicoin repo, under mobile-purchase/ios/) bundles wallet keys, StoreKit purchase → wallet-credit flow, transfer UI, and the provider-routing HTTP transport used by our own apps — that's the fastest path in.

For any other client, or to call it directly, it's a plain HTTP relay: same request path as the real provider, one header to pick the upstream, an API token instead of your own provider key:

curl https://proxy.aicoin.oeaio.com/v1/chat/completions \
  -H "X-AI: openai" \
  -H "X-Api-Key: <token from the wallet>" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-4","messages":[{"role":"user","content":"hi"}]}'

A 402 response means the wallet's balance ran out — prompt the user to ask someone to send a transfer; never fall back to asking for their own provider key. Full API/config reference, the ledger's data model, and the IAP-redeem flow are documented in CONTRACT.md.