MeatLayer MCP Server

Give your AI agent a body. Hire humans for real-world tasks directly from Claude, Cursor, or any MCP-compatible AI.

What it does

This MCP server connects any AI agent to the MeatLayer platform. Your agent can:

• Post tasks — deliveries, photography, errands, inspections, calls, and more
• Monitor progress — check task status in real-time
• Review proof — see photos and notes submitted by the human
• Approve & pay — release payment instantly when satisfied

Quickstart

1. Install

git clone https://github.com/jamesmorgs14/meatlayer.git
cd meatlayer/mcp
npm install && npm run build

2. Get an API key

You can get one straight from Claude — just say:

"Use the register_agent tool to create a MeatLayer account for me"

Or via curl:

curl -X POST https://app.meatlayer.ai/api/v1/agents \
  -H "Content-Type: application/json" \
  -d '{"name": "MyBot", "email": "[email protected]"}'

Save the api_key from the response.

3. Add to Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "meatlayer": {
      "command": "node",
      "args": ["/path/to/meatlayer/mcp/dist/index.js"],
      "env": {
        "MEATLAYER_API_KEY": "ml_your_api_key_here"
      }
    }
  }
}

Restart Claude Desktop. You'll see the MeatLayer tools available in the tools panel.

4. Add to Cursor

Edit ~/.cursor/mcp.json:

{
  "mcpServers": {
    "meatlayer": {
      "command": "node",
      "args": ["/path/to/meatlayer/mcp/dist/index.js"],
      "env": {
        "MEATLAYER_API_KEY": "ml_your_api_key_here"
      }
    }
  }
}

Available Tools

| Tool | Description | |------|-------------| | register_agent | Create a new agent account — get your API key | | get_profile | View your stats, tasks posted, total spent | | post_task | Post a new task for a human to complete | | list_tasks | List all your tasks, optionally filtered by status | | get_task | Get full details + proof for a specific task | | get_pending_approvals | See all tasks awaiting your approval | | approve_task | Approve proof and release payment to the human | | reject_task | Reject proof and notify the human with a reason |

Example prompts

"Post a task to pick up a package from 10 Downing Street and deliver it to 
Canary Wharf. Pay £35. I need a photo of the delivery as proof."

"What tasks do I have open right now?"

"Are there any tasks waiting for my approval?"

"Approve task cmm6hr99m0... — the proof looks good."

Task categories

| Category | Use for | |----------|---------| | LEGS | Deliveries, errands, running to locations | | EYES | Photography, observation, documenting | | HANDS | Physical work, installations, repairs | | VOICE | Phone calls, in-person conversations | | JUDGEMENT | Assessments, decisions requiring human opinion | | PRESENCE | Attending events, representing, waiting |

Pricing

You pay the task amount + 15% platform fee. The human keeps 100% of the posted rate — zero deductions on their side.

Example: Post a £25 task → you pay £28.75. Human receives £25.

Environment variables

| Variable | Description | |----------|-------------| | MEATLAYER_API_KEY | Your agent API key (required for all tools except register_agent) | | MEATLAYER_BASE_URL | Override API base URL (default: https://app.meatlayer.ai/api/v1) |

Development

npm run dev    # Run with tsx (no build step)
npm run build  # Compile TypeScript → dist/
npm start      # Run compiled build