Integration Docs

POST /<route-alias>/v1/messages
x-api-key: rk_xxx
anthropic-version: 2023-06-01
Content-Type: application/json

{
  "model": "claude-sonnet-4-6",
  "max_tokens": 1024,
  "messages": [
    { "role": "user", "content": "Hello" }
  ]
}

{
  "type": "message",
  "role": "assistant",
  "content": [
    { "type": "text", "text": "Hello! How can I help?" }
  ],
  "usage": {
    "input_tokens": 12,
    "output_tokens": 8,
    "cache_creation_input_tokens": 0,
    "cache_read_input_tokens": 0
  }
}

Select a route from the URL path (for example /claude-fast/v1/messages). The buyer ledger meters five token classes: uncached input, output, cache write (5m), cache write (1h), and cache read. Each class is billed independently and debited from your prepaid USDC balance in real time.

Token classes:
• input      — uncached input tokens (prompt)
• output     — generated output tokens
• cache_w5m  — cache write tokens (5-minute TTL)
• cache_w1h  — cache write tokens (1-hour TTL)
• cache_read — cache read tokens (cache hit)

cost = Σ(tokens_in_class × unit_price_for_class)

Example metering for a cached request:
{
  "uncachedInputTokens": 50,
  "cacheWrite5mTokens": 200,
  "cacheWrite1hTokens": 0,
  "cacheReadTokens": 1500,
  "outputTokens": 120,
  "buyerChargeUsdMicros": 4200
}

Cache reads are dramatically cheaper than uncached input. Apply prompt caching to repeated context to cut costs. Cache writes are billed only once per cache creation.

POST /buyer/rotate-api-key
Authorization: Bearer <session_token>

{
  "item": {
    "buyerId": "buyer_abc",
    "routeApiKey": "rk_new_xxx"
  }
}

The old key is revoked instantly. Copy the new key now — it is shown only once. Your browser session is updated automatically.