Skip to content
TakoAPI
Back to skills

lightning-mcp-server

by roasbeef

Web & Frontend Development
826Downloads reported by clawskills.sh
0Views
0 likesSign in to like this skill
—Not yet rated
Your rating

Brief

Build and configure the MCP server for Lightning Node Connect (LNC)

Description

What This Skill Does

MCP server that connects AI assistants to Lightning Network nodes via Lightning Node Connect (LNC). Uses encrypted WebSocket tunnels through a mailbox relay, requiring only a 10-word pairing phrase from Lightning Terminal, no TLS certificates or macaroons. Read-only by default, exposing 18 tools for querying node state, channels, payments, invoices, peers, and on-chain data.

Access a remote lnd node without exposing TLS certificates, macaroons, or open gRPC ports; the mailbox relay handles all connectivity using only an in-memory pairing phrase.

When to Use It

  • Check Lightning node channel balances and capacity mid-conversation
  • Decode a BOLT11 invoice to verify amount and recipient before paying
  • Review payment history and track a specific payment by hash
  • Inspect peer connections and sample Lightning Network topology
  • Estimate on-chain fees for different confirmation targets

Example Workflow

INPUT

User asks: What is the current balance and channel status of my Lightning node?

AGENT
  1. 1

    Call lnc_connect with the 10-word pairing phrase to open an encrypted tunnel

  2. 2

    Call lnc_get_balance to retrieve on-chain wallet and channel balances

  3. 3

    Call lnc_list_channels to get active channels with capacity and local/remote balances

  4. 4

    Call lnc_disconnect to close the tunnel

  5. 5

    Return a summary of balances and channel status to the user

OUTPUT

Node has 0.05 BTC on-chain, 3 active channels totaling 0.12 BTC capacity, with 0.08 BTC local balance available to send

Requirements

  • Lightning Terminal (litd) installed and running on the target lnd node to generate pairing phrases

Installation

clawhub install roasbeef/lightning-mcp-server

Run the CLI command, or paste the prompt into Claude Code, Cursor, Codex or whichever coding agent you use.

Links