MCP server for LightningProx — pay-per-request AI access via Bitcoin Lightning. Supports vision/multimodal.
lightningprox-mcp
以下摘录自该仓库的 README(片段) · 在 GitHub 读完整 README →
MCP server for LightningProx — pay-per-request AI via Bitcoin Lightning. No accounts, no API keys. Load a spend token, start querying.
Install
What LightningProx Is
LightningProx is an AI gateway that accepts Bitcoin Lightning payments instead of API keys. You load a prepaid spend token, pass it in the X-Spend-Token header, and each request is deducted from your balance in sats. No signup, no monthly plan, no credentials to manage.
**19 models across 5 providers:**
| Provider | Models |
|----------|--------|
| Anthropic | claude-opus-4-6, claude-sonnet-4-6, claude-haiku-4-5 |
| OpenAI | gpt-4o, gpt-4-turbo, gpt-4o-mini |
| Together.ai | llama-4-maverick, llama-3.3-70b, deepseek-v3, mixtral-8x7b |
| Mistral | mistral-large-latest, mistral-medium, mistral-small, codestral, devstral, magistral |
| Google | gemini-2.5-flash, gemini-2.5-pro |
**Vision / multimodal:** Pass image_url directly in your request. URL mode only — no base64 encoding required.
Setup
Claude Desktop
**Config location:**
- macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
- Windows: %APPDATA%\Claude\claude_desktop_config.json
- Linux: ~/.config/claude/claude_desktop_config.json
Claude Code
Tools
| Tool | Description |
|------|-------------|
| ask_ai | Send a prompt to any model, authenticated via spend token. Pass model to select (e.g. gemini-2.5-flash, mistral-large-latest, claude-sonnet-4-6). |
| ask_ai_vision | Send a prompt with an image URL for multimodal analysis |
| check_balance | Check remaining sats on a spend token |
| list_models | List available models with per-call pricing |
| get_pricing | Estimate cost in sats for a given model and token count |
| get_invoice | Generate a Lightning invoice to top up a spend token |
Spend Token Auth
Every request authenticates via the X-Spend-Token header:
For vision requests, include image_url in the message content — no base64 needed:
Getting a Spend Token
1. Call get_invoice (or ask_ai without a token) to receive a Lightning invoice
2. Pay the invoice from any Lightning wallet
3. Your spend token is returned — use it for all subsequent requests until balance runs out
Endpoints
| Endpoint | Description |