This server provides 26 paid tools for Bitcoin mempool intelligence and sovereign on-premise AI inference, with automatic payment via Lightning (L402) or USDC on Base (x402).
Bitcoin Mempool Intelligence (from a real Bitcoin Core full node):
get_fees — Fee estimates (next-block, 30-min, 1-hour) with trend analysis (10 sats)
get_whales — Whale alerts for large transfers, consolidations, and distributions in the mempool (40 sats)
get_mempool_state — Live mempool state: tx count, byte size, and fee histogram (5 sats)
get_mempool_history — Rolling 24h snapshots of mempool size and fees (50 sats)
predict_fees — EMA-based fee predictions with confidence scores and direction (rising/falling) (50 sats)
get_recent_blocks — Recent block data: heights, tx counts, and timestamps (10 sats)
tx_status — Look up any Bitcoin transaction by txid for confirmations, fee, block height, and mempool ETA (15 sats)
fee_quote — Exact fee quote for a planned transaction given inputs, outputs, and target blocks (15 sats)
Sovereign AI Inference (Qwen3 on dedicated GPU, no third-party APIs, no data leakage):
ai_summarize — Summarize text at short/medium/long lengths (20 sats)
ai_sentiment — Sentiment analysis with a confidence score (10 sats)
ai_translate — Translate between 100+ languages (20 sats)
ai_grammar — Fix grammar, spelling, and punctuation (15 sats)
ai_code_review — Review code for bugs, security, and performance issues (50 sats)
ai_code_generate — Generate code from natural-language specifications (40 sats)
ai_extract — Extract structured data (entities, contacts, dates, or custom schema) from text (25 sats)
ai_classify — Classify text into user-provided categories (10 sats)
ai_rewrite — Rewrite text in a different style or tone (15 sats)
ai_explain — Explain any topic at ELI5, beginner, intermediate, or expert level (25 sats)
ai_agent — General-purpose AI agent for any text task (30 sats)
ai_scrape — Scrape and extract clean readable text from any URL (15 sats)
ai_embed — Generate 768-dim embedding vectors for RAG (single or batch) (2 sats)
ai_vision — Visual question answering over an image URL (40 sats)
ai_ocr — Extract all visible text from an image URL (25 sats)
ai_transcribe_table — Convert a table image into structured JSON rows (35 sats)
ai_longform — Summarize huge documents (~50K words) in a single flat-fee call (75 sats)
ai_research — Multi-step web research with cited synthesis (100 sats)
Payment options: Lightning Network sats (L402), USDC on Base (x402), or a pre-funded API key (BITCOIN_BENJI_API_KEY). Pricing ranges from 2–100 sats (or USD equivalent) per call.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@@bitcoinbenji/mcp what are current Bitcoin fees?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
MCP server for Bitcoin Benji API — 26 paid tools for Bitcoin mempool intelligence + sovereign on-prem AI inference.
No third-party APIs. No data leakage. Dual payment rails: Lightning sats (L402) or USDC on Base (x402) — with automatic payment.
Set X402_PRIVATE_KEY to an EVM private key holding USDC on Base and every tool
call pays for itself automatically — no Lightning wallet, no API key, no manual
steps. The server signs an EIP-3009 USDC authorization for each 402 challenge
and retries; gas is sponsored by the facilitator, funds go directly seller-side.
{
"mcpServers": {
"bitcoinbenji": {
"command": "bitcoinbenji-mcp",
"env": { "X402_PRIVATE_KEY": "0x<key holding USDC on Base>" }
}
}
}Related MCP server: @powforge/captcha-mcp
Mempool intelligence (from a real Bitcoin Core full node):
get_fees — fee estimates with trend
get_whales — whale/consolidation alerts
get_mempool_state — live mempool state
get_mempool_history — 24h rolling snapshots
predict_fees — EMA-based prediction
get_recent_blocks — recent blocks
tx_status — lookup any txid
fee_quote — exact fee for a planned tx
AI inference (Qwen3 on dedicated GPU, no third-party APIs):
ai_summarize, ai_sentiment, ai_translate, ai_grammar
ai_code_review, ai_code_generate, ai_extract, ai_classify
ai_rewrite, ai_explain, ai_agent, ai_scrape
ai_embed, ai_vision, ai_ocr, ai_transcribe_table
ai_longform (~50K-word docs flat-fee), ai_research (multi-step + cited)
Pricing: 2–100 sats per call (or the USD-equivalent in USDC: 10 sats = $0.01). See api.bitcoinbenji.com/pricing.
npm install -g @bitcoinbenji/mcpEdit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"bitcoinbenji": {
"command": "bitcoinbenji-mcp"
}
}
}Restart Claude. Ask: "What are current Bitcoin fees?" — Claude will call get_fees, hit a 402 challenge, return a Lightning invoice. Pay it with any wallet, paste the preimage back, and Claude completes the call.
Same idea — point your MCP-compatible client at the bitcoinbenji-mcp binary.
Pre-fund a key once, then every call auto-debits in the background:
# Get a key (one-time): visit https://api.bitcoinbenji.com or curl /api/key/create
# Then:
export BITCOIN_BENJI_API_KEY="bb_live_..."
bitcoinbenji-mcpOr in Claude Desktop config:
{
"mcpServers": {
"bitcoinbenji": {
"command": "bitcoinbenji-mcp",
"env": { "BITCOIN_BENJI_API_KEY": "bb_live_..." }
}
}
}Most AI inference APIs require a credit card, an account, and a SaaS subscription. This one runs on a sovereign solar-powered AMD Strix Halo box, accepts payment in sats over Lightning, and never leaks your prompts to OpenAI/Anthropic/Google.
github.com/ocbenji/bitcoinbenji-mcp
MIT
General-purpose AI agent — handles any text task. [30 sats per call]
Inputs: task (required), context, preimage, macaroon.
Classify text into your provided categories. [10 sats per call]
Inputs: text (required), categories (required), preimage, macaroon.
Generate code from a natural-language spec. [40 sats per call]
Inputs: prompt (required), language, context, preimage, +1 more.
Code review for bugs, security, and performance issues. [50 sats per call]
Inputs: code (required), language, preimage, macaroon.
768-dim embedding vector for RAG (single text or batch). [2 sats per call]
Inputs: text, texts, preimage, macaroon.
Explain any topic at the requested level. [25 sats per call]
Inputs: topic (required), level, preimage, macaroon.
Extract structured data from text (entities/contacts/dates/custom schema). [25 sats per call]
Inputs: text (required), schema, custom_fields, preimage, +1 more.
Fix grammar, spelling, and punctuation. [15 sats per call]
Inputs: text (required), preimage, macaroon.
Summarize huge documents (~50K words) in one flat-fee call. [75 sats per call]
Inputs: text, url, focus, max_length, +2 more.
Extract all visible text from an image URL. [25 sats per call]
Inputs: image_url (required), preimage, macaroon.
Multi-step web research + cited synthesis. [100 sats per call]
Inputs: question (required), max_sources, preimage, macaroon.
Rewrite text in a different style/tone. [15 sats per call]
Inputs: text (required), style (required), preimage, macaroon.
Scrape & extract clean readable text from any URL. [15 sats per call]
Inputs: url (required), preimage, macaroon.
Sentiment analysis with confidence score. [10 sats per call]
Inputs: text (required), preimage, macaroon.
Summarize text (short/medium/long). [20 sats per call]
Inputs: text (required), max_length, preimage, macaroon.
Convert a table image into JSON rows. [35 sats per call]
Inputs: image_url (required), preimage, macaroon.
Translate between 100+ languages. [20 sats per call]
Inputs: text (required), target (required), source, preimage, +1 more.
Vision QA over an image URL (Qwen3-VL, sovereign). [40 sats per call]
Inputs: image_url (required), question, preimage, macaroon.
Exact fee quote for a planned tx: {inputs, outputs, target_blocks, input_type}. [15 sats per call]
Inputs: inputs (required), outputs (required), target_blocks (required), input_type, +2 more.
Bitcoin fee estimates with trend analysis (next-block, 30-min, 1-hour) from a full node mempool. [10 sats per call]
Inputs: preimage, macaroon.
Rolling 24h snapshots of mempool size and fees. [50 sats per call]
Inputs: preimage, macaroon.
Live Bitcoin mempool state: tx count, byte size, fee histogram. [5 sats per call]
Inputs: preimage, macaroon.
Recent Bitcoin block data: heights, tx counts, timestamps. [10 sats per call]
Inputs: preimage, macaroon.
Recent whale alerts: large transfers, consolidations, distributions in the mempool. [40 sats per call]
Inputs: since_seconds, preimage, macaroon.
EMA-based fee prediction with confidence + direction (rising/falling). [50 sats per call]
Inputs: preimage, macaroon.
Lookup a Bitcoin transaction by txid: confirmations, fee, block height, mempool ETA. [15 sats per call]
Inputs: txid (required), preimage, macaroon.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 26 tools
All AI tools have distinct purposes (e.g., classify, summarize, translate) and are clearly delineated by the 'ai_' prefix. The Bitcoin tools are also distinct from each other and from the AI tools, leaving no ambiguity.
The AI tools consistently use 'ai_' prefix, but the Bitcoin tools mix 'get_' (get_fees, get_mempool_state) with other patterns (fee_quote, predict_fees), creating an inconsistent overall naming convention.
26 tools is on the high side but reasonable given the dual domain (18 AI + 8 Bitcoin). Each tool appears justified, and the count is not excessive for the scope.
The AI toolset covers many common tasks (generation, review, embedding, OCR, research, etc.). The Bitcoin toolset covers mempool, fees, and transaction status, though it lacks some operations like broadcasting.
No-KYC VPS + per-second code sandboxes AI agents run over MCP — 75 tools, SDK, pay BTC/XMR/USDC/USDT
Bitcoin full node data via MCP: fees, mempool, txs, whales, SEC. $0.001/call USDC on Base.
Bitcoin MCP: fees, tx proofs, address checks. Pay per call by x402, Lightning or a prepaid pass.
Hosted MCP server for live Bittensor chain reads and self-custodial on-chain writes.
