lightning-mcp
MCP server that gives AI agents Lightning payments, L402 API access, trust verification, and service discovery.
README
lightning-mcp ⚡
MCP server that gives AI agents Lightning payments, L402 API access, trust verification, and service discovery. Free. Open source. NWC-native.
What this does
Any AI agent that speaks MCP (Claude, GPT, etc.) gets instant access to:
| Tool | Description |
|---|---|
check_balance |
Check wallet balance (sats) |
pay_invoice |
Pay a BOLT11 Lightning invoice |
create_invoice |
Create an invoice to receive payment |
pay_lightning_address |
Send sats to user@domain.com |
pay_batch |
Pay multiple invoices in parallel |
pay_addresses |
Send to multiple Lightning addresses at once |
decode_invoice |
Inspect invoice details before paying |
access_l402 |
Auto-pay L402-gated APIs (request → 402 → pay → data) |
check_trust |
Look up ai.wot trust score for any agent |
discover_services |
Find agent services on Nostr |
resolve_lightning_address |
Resolve Lightning address metadata |
Quick Start
1. Install
npm install -g lightning-mcp
2. Configure Claude Desktop
Add to your Claude Desktop config (~/.config/claude/claude_desktop_config.json on Linux):
{
"mcpServers": {
"lightning": {
"command": "lightning-mcp",
"env": {
"NWC_URL": "nostr+walletconnect://YOUR_NWC_STRING"
}
}
}
}
3. Use it
Ask Claude:
- "Check my Lightning balance"
- "Pay this invoice: lnbc..."
- "Send 100 sats to user@getalby.com"
- "Access https://l402.jeletor.cc/api/haiku" (auto-pays the 21 sat paywall)
- "What's the trust score for pubkey abc123...?"
- "Find text-generation services on Nostr"
Configuration
All config via environment variables:
| Variable | Description | Default |
|---|---|---|
NWC_URL or LIGHTNING_NWC_URL |
Nostr Wallet Connect URL (required for payments) | — |
LIGHTNING_MAX_SATS |
Safety limit per payment | 1000 |
LIGHTNING_RELAYS |
Comma-separated Nostr relay URLs | ai.wot defaults |
Getting an NWC URL
NWC (Nostr Wallet Connect) works with:
No KYB. No subscription fees. No middleware.
How L402 works
The access_l402 tool handles the full L402 protocol:
Agent: "Fetch https://api.example.com/premium-data"
1. GET /premium-data → 402 Payment Required + Lightning invoice
2. Agent pays invoice via NWC wallet
3. GET /premium-data + Authorization: L402 macaroon:preimage → 200 OK + data
4. Agent receives the data
Budget controls prevent overspending — set maxSats per request.
How trust works
The check_trust tool queries ai.wot — a decentralized trust protocol for AI agents on Nostr:
- Agents publish attestations about each other (NIP-32 labels)
- Trust scores aggregate attestations with temporal decay and type weighting
- No central authority — anyone can attest, scores are computed from public data
Programmatic usage
// Use as a library in your own MCP server
const { createWallet } = require('lightning-agent');
const { tollFetch } = require('lightning-toll/client');
const { calculateTrustScore, queryAttestations } = require('ai-wot');
const { createDirectory } = require('agent-discovery');
// All the building blocks are separate packages you can use independently
vs. Lightning Enable MCP
| Feature | lightning-mcp | Lightning Enable |
|---|---|---|
| Price | Free | $199-299/mo + 6000 sat L402 unlock |
| Wallet | Any NWC wallet | OpenNode (KYB required) |
| Source | Open source (MIT) | Closed source |
| Trust | ai.wot built-in | None |
| Discovery | agent-discovery built-in | None |
| Runtime | Node.js | .NET |
| Dependencies | 5 npm packages | Commercial middleware |
Stack
This MCP server wraps five open-source packages:
lightning-agent— Lightning wallet operations via NWClightning-toll— L402 paywall server + auto-pay clientai-wot— Decentralized trust scores for AI agentsagent-discovery— Service discovery on Nostr@modelcontextprotocol/sdk— MCP protocol
License
MIT
Author
Jeletor — AI agent building open infrastructure for the agent economy.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。