Usage and credits

Per-token rates and the standardized Pokee Isaac tiers (pokee-isaac-high-reasoning, pokee-isaac) are documented on the Models and pricing page. This page covers the /v1/usage endpoint, the rate-limit buckets, and the event types that show up in your audit log.

GET /v1/usage

Returns rolling rate-limit usage and credit balance.

{
  "rate_limit": {
    "caller_id": "default",
    "buckets": {
      "session_turn": {
        "windows": [
          {"name": "hour", "duration_seconds": 3600,  "turns": 5, "max_turns": 500,  "tokens": 31137, "max_tokens": 100000000,  "enabled": true},
          {"name": "day",  "duration_seconds": 86400, "turns": 5, "max_turns": 5000, "tokens": 31137, "max_tokens": 1000000000, "enabled": true}
        ]
      },
      "response": {
        "windows": [
          {"name": "hour", "duration_seconds": 3600,  "turns": 0, "max_turns": 500,  "tokens": 0, "max_tokens": 100000000,  "enabled": true},
          {"name": "day",  "duration_seconds": 86400, "turns": 0, "max_turns": 5000, "tokens": 0, "max_tokens": 1000000000, "enabled": true}
        ]
      }
    }
  },
  "credits": {
    "enabled": true,
    "mode": "soft",
    "balance": 99961.99
  }
}

balance is the credits remaining for your tenant. The exact debit per call depends on what the agent does — read balance before and after a call if you want to see the precise impact.

Rate-limit buckets

Each endpoint surface has its own independent rate-limit budget. A runaway loop hitting one bucket can't drain the others. Buckets:

All buckets share the same max_turns / max_tokens caps in this release; per-bucket overrides are on the roadmap.

Event types

Each chargeable event in your audit log is tagged with one of these event_type values:

Modes

• soft — debits recorded, never reject.
• hard — debits recorded AND a turn is rejected with 402 Payment Required when balance falls below the per-turn minimum.

If you hit a 402, contact Pokee for a top-up.