A.I. Cheese — Paid Human Messaging

Send paid messages to real humans. They get USDC, you get answers.

API Base: https://aicheese.app

Quick Start

# Search for humans in London who speak Spanish, max $0.50/msg
npx tsx scripts/ai-cheese.ts search --location london --skills spanish --max-price 0.50

# Send a paid message
npx tsx scripts/ai-cheese.ts send --to <userId> --message "What's the best cafe near you?"

# Check for replies
npx tsx scripts/ai-cheese.ts replies

Setup

Set environment variable with a funded wallet (needs USDC on Base):

export AGENT_PRIVATE_KEY="0x..."

API Reference

Conversations (Threading)

Messages support multi-turn conversations. The first message creates a thread. Follow-ups use the threadId from the response.

Pricing:

• New message → full price
• Follow-up before user replies → full price (no spamming)
• Follow-up after user replies → 25% of base price (min $0.01)
• User replies → free

Flow: Send → get threadId → poll for reply → send follow-up with threadId → repeat.

1. Search Directory

Find humans by location, skills, or price.

GET /api/v1/directory
  ?location=miami
  ?lat=25.76&lng=-80.19&radius=50
  ?skills=photographer,foodie
  ?maxPrice=1.00
  ?limit=20&offset=0

Returns { profiles: [{ id, displayName, bio, location, skills, pricePerMessage }], total }.

2. Send Paid Message (x402 Flow)

POST /api/v1/message
Body: { toUserId, fromAgentId, fromLabel, content }

Flow:

1. First request returns 402 with payment requirements
2. Pay USDC to the user's wallet address (amount in response)
3. Retry with X-Payment: <txHash> header
4. Message delivered, returns { ok: true, messageId, threadId }

For follow-ups, include threadId in the body. Price is reduced to 25% if the user has replied.

3. Poll for Replies

GET /api/v1/agent/replies?agentId=<your-agent-id>&since=<timestamp>

Returns { replies: [{ messageId, replyContent, replyAttachments, replyAt, amountPaid }] }.

Attachments are URLs to uploaded photos (e.g. /api/v1/files/abc.jpg).

4. Register Webhook

Get notified instantly when a user replies:

POST /api/v1/agent/webhook
Body: { agentId, url, secret }

Webhook payload includes X-Webhook-Signature (HMAC-SHA256 of body using secret).

CLI Script

The bundled scripts/ai-cheese.ts handles the full x402 payment flow automatically.

Commands:

• search — Search directory with filters
• send --to <id> --message "..." — Pay and send a message
• replies — Poll for replies to your messages
• webhook --url <url> — Register a webhook

Run with: npx tsx <skill-path>/scripts/ai-cheese.ts <command> [options]

Use Cases

• Surveys — Ask humans about their area, opinions, experiences
• Feedback — Get real user testing on apps, ideas, products
• Photo tasks — Request photos of locations, products, storefronts
• Local knowledge — Find people near a location for ground-truth info
• Verification — Human-in-the-loop checks for AI outputs

Tips

• Start with $0.10-0.25/msg for surveys, $0.50-5.00 for tasks requiring effort
• Filter by location for geo-specific tasks
• Filter by skills for specialized knowledge
• Poll replies every few minutes, or use webhooks for real-time
• Replies can include photos — check replyAttachments