Per-Key Plans
Each pool subscription mints its own unlimited API key. To isolate different users, clients, or applications, subscribe to multiple seats (one key per seat) — each key is independent, so one runaway client can’t starve the others (each key allows 1 concurrent request).
Use cases
Section titled “Use cases”- SaaS platform — Give each customer their own API key with appropriate limits
- Team management — Different keys for dev, staging, production
- Client billing — Track usage per client with separate keys
Creating keys with plans
Section titled “Creating keys with plans”From the dashboard
Section titled “From the dashboard”- Go to API Keys
- Click Create Key
- Set a name (e.g.
client-acme) - Select a subscription to attach the key to
Via API
Section titled “Via API”Subscribe to a pool, reserving one or more seats (quantity) across your chosen time blocks. Each seat becomes an independent subscription you can mint a key from:
curl -X POST https://api.cheapestinference.com/api/pools/kimi26/subscribe \ -H "Authorization: Bearer mk_your_management_key" \ -H "Content-Type: application/json" \ -d '{"blocks": ["asia","europe","americas"], "quantity": 3}'Then create a key for each subscription (one active key per subscription):
curl -X POST https://api.cheapestinference.com/api/keys/subscription \ -H "Authorization: Bearer mk_your_management_key" \ -H "Content-Type: application/json" \ -d '{"name": "client-acme", "subscriptionId": "pledge_uuid"}'You need an active pool subscription to create keys. (Autonomous agents can subscribe and pay with USDC in one call via POST /api/agent/subscribe-pool — see the x402 guide.)
Rate limits per key
Section titled “Rate limits per key”Every pool key is unlimited during your reserved time blocks — no RPM/TPM budget cap. The only ceiling is 1 concurrent request per key, so to run requests in parallel (or isolate clients), reserve more seats and use one key each.
| Pool key | |
|---|---|
| Throughput during reserved hours | Unlimited (no budget cap) |
| Concurrent requests per key | 1 |
| Outside reserved hours | Requests return 429 |
Keys are independent — one key’s concurrency does not affect other keys.
Key lifecycle
Section titled “Key lifecycle”- Create — Key is generated with the plan’s rate limits
- Active — The platform validates the key on each request and enforces limits
- Disable/Delete — Key is immediately revoked
When a subscription expires, its keys are automatically revoked by a background job. Other subscriptions and their keys are not affected.
Tracking keys
Section titled “Tracking keys”View all your keys and their subscription status:
curl https://api.cheapestinference.com/api/usage \ -H "Authorization: Bearer mk_your_management_key"Or view them in the dashboard.