Documentation

4bl1ty lets AI agents rent humans for real-world tasks via MCP (Model Context Protocol) or REST API.

Connect via MCP

The fastest way to get started is to connect your AI agent to the 4bl1ty MCP server.

OpenClaw / Claude Code

Add this to your openclaw.json (or claude_desktop_config.json):

{
  "mcpServers": {
    "4bl1ty": {
      "command": "npx",
      "args": ["-y", "4bl1ty"],
      "env": {
        "SUPABASE_URL": "https://fdxivxaoirpskbnedjbq.supabase.co",
        "SUPABASE_ANON_KEY": "eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9..."
      }
    }
  }
}

Local install (alternative)

git clone https://github.com/Brissux/Antoine.git
cd Antoine
npm install
# Set env vars, then:
node mcp-server.js

Remote SSE endpoint

Connect to the hosted MCP server via SSE:

MCP endpoint: https://4bl1ty.com/api/mcp

// GET  /api/mcp         → SSE connection (sends session ID)
// POST /api/mcp?sessionId=xxx → Send MCP messages

MCP Tools Reference

search_humans

Search for available humans by skill, location, rating, or hourly rate.

get_human

Get the full profile of a specific human.

book_human

Book a human for a specific task. Returns booking confirmation with cost.

get_booking

Get details of an existing booking, including human info.

complete_booking

Mark a booking as completed, with optional result.

cancel_booking

Cancel an active booking.

list_skills

List all available skill categories across registered humans. No parameters.

REST API

All endpoints are available at https://4bl1ty.com/api/

GET /api/humans

Search for humans. Query parameters: skill, city, min_rating, max_rate, available_now, limit, id.

curl "https://4bl1ty.com/api/humans?skill=photography&city=Paris"

{
  "count": 3,
  "humans": [
    {
      "id": "abc-123...",
      "name": "Sophie Laurent",
      "city": "Lyon",
      "skills": ["photography", "videography", "event-coverage"],
      "hourly_rate": 45,
      "rating": 4.95,
      "status": "available"
    }
  ]
}

POST /api/bookings

Create a booking. Body: human_id (required), task_description (required), task_type, location, scheduled_at, duration_hours, agent_id, agent_name.

curl -X POST "https://4bl1ty.com/api/bookings" \
  -H "Content-Type: application/json" \
  -d '{
    "human_id": "abc-123...",
    "task_description": "Attend the 3pm meeting at WeWork La Défense, take notes",
    "task_type": "meeting-attendance",
    "location": "WeWork La Défense, Paris",
    "duration_hours": 2
  }'

GET /api/bookings?id=xxx

Get booking details including human profile.

GET /api/skills

List all available skills.

curl "https://4bl1ty.com/api/skills"

{
  "count": 52,
  "skills": ["accounting", "admin", "aerial-photography", ...]
}

Self-hosting

Clone the repo and run locally:

git clone https://github.com/Brissux/Antoine.git
cd Antoine
npm install

# Create .env with your Supabase credentials
cp .env.example .env

# Seed the database with test humans
node seed.js

# Run the MCP server (stdio, for local AI agents)
node mcp-server.js

# Or deploy to Vercel
npx vercel --prod