Hosted API

Don't want to install the SDK? Use InferShrink as a hosted API. Pay per call with USDC on Base chain — no API key, no account, no signup.

Base URL

https://musashi-runtime.musashi-labs.workers.dev

Endpoints

How x402 Payment Works

Every paid endpoint returns HTTP 402 Payment Required with a Payment-Required header containing a base64-encoded JSON object. Here's the flow:

1. Make a request

curl -X POST https://musashi-runtime.musashi-labs.workers.dev/classify \
  -H "Content-Type: application/json" \
  -d '{"prompt": "What is 2+2?"}'

2. Get the 402 response

HTTP/2 402
Payment-Required: eyJ4NDAyVmVyc2lvbi...  (base64)

{
  "error": "Payment required",
  "price": "$0.001 USDC",
  "network": "eip155:8453",
  "description": "InferShrink: Classify prompt complexity"
}

3. Decode the payment instructions

Base64-decode the Payment-Required header to get:

{
  "x402Version": 2,
  "resource": {
    "url": "https://musashi-runtime.musashi-labs.workers.dev/classify",
    "description": "InferShrink: Classify prompt complexity",
    "mimeType": "application/json"
  },
  "accepts": [{
    "scheme": "exact",
    "network": "eip155:8453",
    "amount": "1000",
    "payTo": "0xbdF05e6E011a7E032821a6026A9d47D10C8b9848",
    "maxTimeoutSeconds": 300,
    "asset": "0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913",
    "extra": { "name": "USD Coin", "version": "2" }
  }]
}

4. Pay on Base chain

Send USDC to the payTo address on Base (chain ID 8453):

• Token: USDC (0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913)
• Network: Base (eip155:8453)
• Amount: 1000 = $0.001 USDC (6 decimals)
• Recipient: 0xbdF05e6E011a7E032821a6026A9d47D10C8b9848

5. Include payment proof

After the on-chain transaction confirms, re-send your original request with the transaction hash:

curl -X POST https://musashi-runtime.musashi-labs.workers.dev/classify \
  -H "Content-Type: application/json" \
  -H "X-Payment-Tx: 0xabc123..." \
  -d '{"prompt": "What is 2+2?"}'

Python Client

The SDK includes a hosted API client:

from infershrink.hosted import InferShrinkClient, PaymentRequired

client = InferShrinkClient()

try:
    result = client.classify("What is 2+2?")
    print(result)  # {"complexity": "SIMPLE", ...}
except PaymentRequired as e:
    print(e.details)
    # {"price": "$0.001 USDC", "network": "eip155:8453", ...}
    # Your agent can parse this and pay automatically

For Agent Developers

If you're building an AI agent that needs to classify prompt complexity or route models, the x402 flow is designed for you:

1. Your agent calls /classify or /route
2. Gets 402 with structured payment instructions
3. Agent's wallet signs a USDC transfer on Base
4. Agent retries with tx proof — gets the classification result
5. No API keys to manage, no accounts to create, no rate limits to hit

API Reference

POST /classify

// Request
{"prompt": "What is 2+2?"}

// Response (after payment)
{
  "complexity": "SIMPLE",
  "reason": "short prompt, no complex keywords",
  "signals": {
    "estimated_tokens": 6,
    "complex_keywords": [],
    "code_blocks": 0
  }
}

POST /route

// Request
{"model": "gpt-4o", "prompt": "What is 2+2?"}

// Response (after payment)
{
  "original_model": "gpt-4o",
  "routed_model": "gpt-4o-mini",
  "was_downgraded": true,
  "complexity": "SIMPLE",
  "estimated_savings_pct": 95
}

POST /subscribe

// Request (empty body)
{}

// Response (after $1.01 USDC payment)
{
  "credits": 1000,
  "expires_at": "2026-03-10T17:00:00Z",
  "token": "isk_abc123..."
}

// Use token in subsequent requests:
// -H "Authorization: Bearer isk_abc123..."