Chutes AI Proxy (x402)

Build a pay-per-inference proxy for Bittensor Chutes (Subnet 64). Accept USDC payments on Base and forward requests to Chutes' decentralized AI network.

When to Use This Skill

Use this skill when the user wants to:

• Build an AI inference API that accepts crypto payments
• Wrap Chutes/Bittensor with their own pricing
• Create a pay-per-request AI service without subscriptions
• Monetize AI access using stablecoins
• Set up a Bittensor-powered AI endpoint

How to Respond

Quick Start

Node.js / TypeScript

npx @primersystems/x402 create chutes-proxy my-ai-proxy
cd my-ai-proxy
npm install
cp .env.example .env
# Edit .env with your Chutes API key and wallet address
npm run dev

Python

pip install primer-x402
x402 create chutes-proxy my-ai-proxy
cd my-ai-proxy
pip install -r requirements.txt
cp .env.example .env
# Edit .env with your Chutes API key and wallet address
uvicorn main:app --reload

How It Works

User Request
     |
     v
[Your Proxy] -- No payment? --> Return 402 + price
     |
     v
User signs USDC payment (gasless)
     |
     v
[Your Proxy] -- Verify payment --> [Primer Facilitator]
     |
     v
[Chutes API] -- Forward request --> Bittensor Subnet 64
     |
     v
AI Response returned to user

You pre-pay Chutes with your API credits. Users pay YOU in USDC. You keep the margin.

Configuration

After scaffolding, edit .env:

Getting a Chutes API Key

1. Go to chutes.ai
2. Sign up / connect wallet
3. Subscribe to a tier ($3/month base)
4. Generate API key (starts with cpk_)

Deployment

Cloudflare Workers (TypeScript - Free)

wrangler login
wrangler secret put CHUTES_API_KEY
wrangler secret put WALLET_ADDRESS
npm run deploy

Docker (Python)

docker build -t chutes-proxy .
docker run -p 8000:8000 --env-file .env chutes-proxy

Other Platforms

• fly.io: fly launch && fly secrets set CHUTES_API_KEY=xxx
• Railway/Render: Connect repo, set env vars in dashboard
• Vercel Edge: Build and deploy TypeScript version

API Endpoints

Your proxy exposes:

Supported Models

Any model on Chutes, including:

• deepseek-ai/DeepSeek-V3
• Qwen/Qwen3-235B-A22B
• meta-llama/Llama-3.1-70B-Instruct
• meta-llama/Llama-3.1-8B-Instruct

See chutes.ai for the full list.

Pricing Strategy

The proxy estimates tokens and charges upfront:

Price = (estimated_tokens / 1000) * PRICE_PER_1K_TOKENS

Set PRICE_PER_1K_TOKENS higher than Chutes' cost to make margin. Example:

• Chutes costs you ~$0.0005/1K tokens
• You charge $0.001/1K tokens
• You keep 50% margin

Limitations

• Streaming not supported - Template doesn't handle stream: true
• Token estimation is approximate - Uses ~4 chars/token heuristic
• Pre-payment only - No post-inference reconciliation

Use Cases

Testing Your Proxy

# Should return 402 Payment Required
curl http://localhost:8787/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"messages":[{"role":"user","content":"Hello"}]}'

# Pay and get response (using x402 CLI)
npx @primersystems/x402 pay http://localhost:8787/v1/chat/completions \
  --max-amount 0.01 \
  --method POST \
  --body '{"messages":[{"role":"user","content":"Hello"}]}'

Links

• Chutes AI: https://chutes.ai
• Bittensor: https://bittensor.com
• x402 Protocol: https://x402.org
• Primer Systems: https://primer.systems
• SDK Documentation: https://x402.org/docs/bittensor.html
• GitHub: https://github.com/primer-systems/x402