MCP Server for Bitcoin Lightning

This repository uses https://github.com/ehallmark/btc-lightning-client for communicating with a local Lightning Network Daemon (lnd) and wraps it in a Model Context Protocol (MCP) server. This MCP server can be used with Agentic systems to interact with the Lightning Network.

Prerequisites

• Python >= 3.10
• uv for package management
• You must have a running Lightning Network Daemon (lnd) running on your machine. For more information on lnd, visit https://github.com/lightningnetwork/lnd.

Running the MCP Server

export LIGHTNING_RPC_PORT=<lnd_port>
export LIGHTNING_CERT_PATH="/path/to/lnd/tls.cert"
export LIGHTNING_MACAROON_PATH=/path/to/bitcoin/simnet/admin.macaroon
export UVICORN_PORT=<some port>
export UVICORN_HOST=<some host>

uv run main.py

Connecting to the MCP Server

This server uses SSE for transport.

With the LangGraph framework, you can connect to it using the following adapter:

from contextlib import asynccontextmanager
from langchain_mcp_adapters.client import MultiServerMCPClient

@asynccontextmanager
async def make_graph():
    async with MultiServerMCPClient(
        {
            'lightning': {
                'url': 'http://localhost:8000/sse',
                'transport': 'sse',
            }
        }
    ) as lightning_mcp:
        workflow = create_workflow(user, lightning_mcp)
        graph = workflow.compile()
        yield graph