Breez MCP Server
Enables Lightning Network wallet operations through the Breez SDK, allowing users to check balances, send/receive payments, create invoices, and manage Bitcoin Lightning transactions via natural language.
README
Breez MCP Server — FastMCP Implementation
A unified MCP server that exposes Lightning functionality through the Breez SDK (Spark implementation) using FastMCP. Supports both stdio and HTTP transport modes.
Prerequisites
- Python 3.11+ (for local development or
uvx) - Docker (optional, for container workflows)
- uv (optional, for ephemeral environments)
- Breez API key which you can request here
Configure Credentials
cp .env.example .env
Edit .env with your secrets. Required variables:
| Variable | Required | Default | Purpose |
|---|---|---|---|
BREEZ_API_KEY |
✅ | – | Breez Spark API key |
BREEZ_MNEMONIC |
✅ | – | 12-word mnemonic controlling the wallet |
BREEZ_NETWORK |
❌ | mainnet |
Set to testnet for sandbox usage |
BREEZ_DATA_DIR |
❌ | ./data |
Wallet storage directory |
BREEZ_TRANSPORT_MODE |
❌ | stdio |
Transport mode: stdio, http, or asgi |
BREEZ_HTTP_HOST |
❌ | 0.0.0.0 |
HTTP server host (HTTP mode only) |
BREEZ_HTTP_PORT |
❌ | 8000 |
HTTP server port (HTTP mode only) |
BREEZ_HTTP_PATH |
❌ | /mcp |
HTTP endpoint path (HTTP mode only) |
Run the Server
Choose the runtime that transport mode that fits your workflow.
STDIO Mode (Default for MCP clients)
For use with Claude Desktop and other MCP clients:
# Local virtualenv
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -r requirements.txt
python -m src.main
# Or with uvx (no persistent venv)
uvx --from . breez-mcp
HTTP Mode (for web API access)
For accessing the MCP server via HTTP API:
# Set environment variable
export BREEZ_TRANSPORT_MODE=http
# Or add to .env file
echo "BREEZ_TRANSPORT_MODE=http" >> .env
# Run the server
python -m src.main
The server will be available at http://localhost:8000/mcp
ASGI Mode (for external ASGI servers)
For deployment with external ASGI servers like Gunicorn:
# Set environment variable
export BREEZ_TRANSPORT_MODE=asgi
# Run with uvicorn
uvicorn src.main:app --host 0.0.0.0 --port 8000
# Or with Gunicorn (production)
gunicorn src.main:app -w 4 -k uvicorn.workers.UvicornWorker
Docker Compose
Run both modes simultaneously:
# STDIO mode
docker compose --profile stdio up -d
docker compose logs -f breez-mcp-stdio
# HTTP mode
docker compose --profile http up -d
docker compose logs -f breez-mcp-http
# Stop
docker compose --profile http down
docker compose --profile stdio down
Docker (direct)
# Build image
docker build -t breez-mcp .
# STDIO mode (default)
docker run --rm \
-e BREEZ_API_KEY="$BREEZ_API_KEY" \
-e BREEZ_MNEMONIC="$BREEZ_MNEMONIC" \
-v $(pwd)/data:/app/data \
breez-mcp
# HTTP mode
docker run --rm -p 8000:8000 \
-e BREEZ_TRANSPORT_MODE=http \
-e BREEZ_API_KEY="$BREEZ_API_KEY" \
-e BREEZ_MNEMONIC="$BREEZ_MNEMONIC" \
-v $(pwd)/data:/app/data \
breez-mcp
To keep STDIN/STDOUT attached for Claude Desktop, add -i to the docker run command.
Claude Desktop Integration
Quick install
mcp install src.main --name "breez-mcp"
Use -f .env or -v KEY=value to supply credentials during installation if desired.
Docker from Claude Desktop
Ensure the image exists (docker build -t breez-mcp .), then configure:
{
"mcpServers": {
"breez": {
"command": "docker",
"args": [
"run", "--rm", "-i",
"-e", "BREEZ_API_KEY",
"-e", "BREEZ_MNEMONIC",
"-e", "BREEZ_TRANSPORT_MODE=stdio",
"-v", "/absolute/path/to/breez-mcp/data:/app/data",
"breez-mcp"
],
"cwd": "/absolute/path/to/breez-mcp",
"env": {
"BREEZ_API_KEY": "${env:BREEZ_API_KEY}",
"BREEZ_MNEMONIC": "${env:BREEZ_MNEMONIC}",
"BREEZ_NETWORK": "mainnet"
}
}
}
}
Docker's -e VAR syntax reads the value of VAR from the environment supplied via the env block.
uvx from Claude Desktop
{
"mcpServers": {
"breez": {
"command": "uvx",
"args": ["--from", ".", "breez-mcp"],
"cwd": "/absolute/path/to/breez-mcp",
"env": {
"BREEZ_API_KEY": "${env:BREEZ_API_KEY}",
"BREEZ_MNEMONIC": "${env:BREEZ_MNEMONIC}",
}
}
}
}
Verification
- Restart Claude Desktop after adding the configuration.
- Run
mcp listto ensure the server registered. - Ask Claude prompts like “Check my wallet balance” or “Create an invoice for 1000 sats” to validate tool routing.
Available Tools
get_balance— comprehensive wallet balance with limits and formatted amountsget_node_info— detailed node information including capabilities and sync statussend_payment— send a Lightning payment with complete transaction detailscreate_invoice— generate a BOLT11 invoice with all invoice datalist_payments— comprehensive payment history with full details
Example Prompts
- "Check my wallet balance"
- "Create an invoice for 1000 sats for coffee"
- "Send payment to lnbc1…"
- "Show me my recent payments"
HTTP API Usage (HTTP Mode)
When running in HTTP mode, you can interact with the MCP server via HTTP requests:
Health Check
curl http://localhost:8000/health
List Available Tools
curl http://localhost:8000/mcp/tools/list
Call a Tool (MCP Protocol)
The HTTP mode follows the MCP protocol over HTTP. You'll need to send properly formatted MCP JSON-RPC requests to http://localhost:8000/mcp.
Example using MCP Inspector or other MCP clients:
{
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"name": "get_balance",
"arguments": {}
},
"id": 1
}
Send to:
curl -X POST http://localhost:8000/mcp \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","method":"tools/call","params":{"name":"get_balance","arguments":{}},"id":1}'
Security Notes
- Never commit
.env; keep secrets in your shell or a secrets manager. - Treat the mnemonic as the wallet’s private key. Rotate immediately if leaked.
- Default network is
mainnet. For experimentation, explicitly setBREEZ_NETWORK=testnet. - When using containers, mount
./datato preserve state between runs and prevent secret leakage in container layers.
Troubleshooting
- Missing environment variables — ensure
.envexists or export the required variables before starting. - SDK connection failures — verify required env vars, try
python list_payments_cli.py --limit 1 --verboseto confirm SDK connectivity, and checkhttp://localhost:8000/healthin HTTP mode. - Claude Desktop cannot find the server — double-check absolute paths in
cwdand restart the application after configuration changes.
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。
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。
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。
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。
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。
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。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
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。