AIKeys

Pay per token for frontier models.

Call frontier models through one OpenAI- and Anthropic-compatible API. You pay per token. There is no monthly plan.

I want API access

Pay per token. Maximum value.

Deposit USDC, create one key, and call a listed model — charges stop when you stop calling.

—total requests

—tokens routed

Top 5 consumer spendingAll time

Loading…

Start building

I have capacity

Turn idle capacity into USDC.

Connect Antigravity, Claude Code, or OpenAI Codex and get credited for the capacity you serve.

—ready to spend

—credited to publishers

Top 5 publisher earningsAll time

Loading…

Become a publisher

Today (UTC)

—requests

—input tokens

— cached

—output tokens

  • Asks stay at or below half the official price.
  • USDC deposits on Solana mainnet.
  • OpenAI and Anthropic compatible.

Upstreams

Three upstreams.

Connect one, set an ask, and wait for traffic.

  • Antigravity

    Curated upstream.

  • Claude Code

    Anthropic-compatible.

  • OpenAI Codex

    OpenAI-compatible.

Product

One key. Two protocols.

A marketplace sits behind the key.

  • One key, both protocols

    Send chat completions, messages, or responses. The same key works for each. Or use the helper.

  • Cheapest live ask

    The proxy chooses the lowest serving ask, then the lowest output ask. It does not place a request above the published cap.

  • Serving capacity

    A request is placed only on a serving offer. If none qualifies, the proxy returns 503 no_capacity. You are not charged for that.

  • Pay per use

    Each settled request draws from your balance. Stop calling and the charges stop.

  • Encrypted credentials

    Publisher credentials are encrypted at rest and used only while a request is sent. The API does not return the raw credential.

  • One retry

    The proxy calls the upstream and retries once. The default key rate limit is 60 requests a minute.

Docs

Get started in one command.

Run the helper, or point a client at the inference base URL. The written steps are on Docs.

Terminal
npx @aikeys-co/helper
curl https://api.aikeys.co/v1/chat/completions \
  -H "Authorization: Bearer $AIKEYS_CO_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"cc/claude-haiku-4-5","messages":[{"role":"user","content":"Hello"}]}'

Pricing

Live asks.

Prices are the current lowest ask per million tokens. A model with no serving publisher is unavailable. You pay the lowest live ask, and that ask is never more than half the official price. A publisher cannot set an ask above that cap.

ModelFamilyAsk inAsk out
Loading prices…

Open pricing

Estimate

What would it cost?

Pick a listed model and drag the token volume. The estimate uses the live asks above and compares them with the official price.

Loading prices…

Accounts

Call, or list capacity.

  1. 1

    One account

    Call models, or list capacity.

  2. 2

    Deposit and call

    Deposit USDC, create one key, and use the samples on Get started.

  3. 3

    Connect and price

    Connect a curated upstream and set an ask at the published cap.

  4. 4

    Meter and credit

    You are credited for each request you serve. The split is in the fee section.

Publishers

Put capacity to work.

You are credited for each request you serve. Payouts are not available in this version. A transfer to your consumer balance is in the signed-in product. It is not a withdrawal.

Example

Connected
a curated upstream
Ask
at the published cap
Earnings
credited on the platform
Payout
not available

Platform

20% on each settled request.

The platform keeps 20% of each settled charge. The publisher is credited 80%.

Start building, or become a publisher.

The same account does both.