tollbooth-mcp

Watch agents pay per call — inside your own AI.

An MCP server that demonstrates the full x402 pay-per-call loop in paper mode: your assistant asks for a quote, gets an HTTP-402-style challenge, pays a simulated toll, and receives the result with a genuine receipt from the live nano-empire-tollbooth escrow lifecycle. No real money moves.

Install

pip install tollbooth-mcp

Claude Desktop / Claude Code

{
  "mcpServers": {
    "tollbooth": { "command": "tollbooth-mcp" }
  }
}

Cursor

{ "mcp": { "servers": { "tollbooth": { "command": "tollbooth-mcp" } } } }

Then ask your assistant: "Use the tollbooth to run a paid call and show me the receipt."

Tools

Monetize your own functions

The real thing is one decorator:

from nano_empire_tollbooth import monetize

@monetize(price_usd=0.01)
def summarize(text: str) -> str:
    return my_llm(text)

• Quickstart: https://neuralempireai.com/docs/quickstart.html
• Live simulator: https://neuralempireai.com/simulator/
• Source: https://github.com/roblambert9/nano-empire-tollbooth

Paper mode is a simulation. No revenue is guaranteed; this is a metering SDK, not an income product. MIT licensed.