Give Claude, ChatGPT, or any custom agent live Bitcoin fee and mempool context via MCP so it can answer practical questions like "should I send now?", "what fee gets me into the next 3 blocks?", and "what will this transaction cost?"
The standard that connects AI to external tools.
Model Context Protocol (MCP) is an open standard that lets AI assistants call external tools during a conversation. Instead of the AI guessing or relying on stale training data, it makes a structured tool call to a live data source and gets back a current response.
An MCP server exposes a set of tools — functions that the AI can invoke with parameters and receive structured results. The AI decides when to call which tool based on the user's question. This is fundamentally different from RAG or prompt stuffing: the AI actively queries live systems rather than searching pre-loaded documents.
bitcoin-mcp is the MCP server for Bitcoin. It wraps every Satoshi API endpoint as an MCP tool, so an AI agent can call get_recommended_fees or get_latest_block the same way it would call a calculator or web search.
Four layers from AI agent to Bitcoin node.
Each layer is a separate package. Install what you need: the API alone for REST access, or the full stack for AI integration.
Real fee questions, real endpoints, current answers.
/fees/recommended and returns fee estimates in sat/vB for high, medium, and low priority, along with a human-readable recommendation./fees/landscape, which analyzes mempool congestion and fee trends to produce a send-now-or-wait recommendation with reasoning./blocks/latest and returns block height, transaction count, total fees, weight utilization, miner, and time since the previous block./mempool and returns congestion level, transaction count, total size in vMB, fee buckets, and the minimum fee to enter the next block./mining and returns current difficulty, estimated hashrate, blocks until the next retarget, and the projected adjustment percentage./fees/estimate-tx with the input/output counts and returns estimated size in vBytes and cost at each fee priority level.Three steps from zero to AI-powered Bitcoin fee decisions.
Point it at your Bitcoin Core node. The API starts at localhost:9332.
$ pip install satoshi-api
$ export BITCOIN_RPC_USER=your_user
$ export BITCOIN_RPC_PASSWORD=your_password
$ satoshi-apiThe MCP server wraps every Satoshi API endpoint as an AI tool call.
$ pip install bitcoin-mcpAdd the MCP server to Claude Desktop, or integrate with any MCP-compatible agent.
// claude_desktop_config.json
{
"mcpServers": {
"bitcoin": {
"command": "bitcoin-mcp",
"env": {"SATOSHI_API_URL": "http://localhost:9332"}
}
}
}For hosted API access without running a node, point at the public endpoint:
"env": {"SATOSHI_API_URL": "https://bitcoinsapi.com"}Fee and transaction queries are sensitive. Keep them private.
When an AI agent queries a third-party API, the API provider sees every request: which addresses were looked up, which transactions were checked, what fee thresholds triggered a send decision. Over time, these queries build a detailed profile of the user's Bitcoin activity.
Self-hosting Satoshi API eliminates this entirely. The AI agent talks to your local API, which talks to your local node. No third party sees the queries. No logs leave your machine.
AI agent sends queries to an external server. The provider sees which addresses, transactions, and fee thresholds the agent queries. Requests are logged with your IP. Query patterns reveal financial behavior over time.
AI agent queries your local API on localhost. Requests never leave your machine. No third-party logs, no IP correlation, no query profiling. Your node, your data, your privacy.
MCP plus fee intelligence is the real differentiator.
Most Bitcoin APIs are designed for direct application use, not AI tool calling. Teams that want to give agents access to Bitcoin data usually end up building their own wrappers, auth handling, and response normalization on top.
Satoshi API and bitcoin-mcp provide this out of the box. One pip install for the API, one for the MCP server, and your agent gets structured tool-call access to live Bitcoin fee and mempool workflows without custom plumbing.
| Capability | Satoshi API | Other Bitcoin APIs |
|---|---|---|
| MCP server available | Yes (bitcoin-mcp) | No |
| AI tool-call access | Native, one install | Build it yourself |
| Self-hosted option | Yes (your node) | Rarely |
| Privacy (no query logging) | Full (localhost) | Provider sees all queries |
| Analyzed responses | Fee recs, congestion scores | Raw data or basic |
| Open source | Apache-2.0 license | Varies (mostly no) |
The combination of Satoshi API + bitcoin-mcp gives you a practical way to give an AI agent direct, structured, tool-call access to a Bitcoin full node. Whether the agent is Claude asking about fees, a GPT-based bot checking mempool congestion, or a custom workflow monitoring mining difficulty, this stack provides the interface.
Two pip installs. Three lines of config. Live fee recommendations and mempool context via MCP.
Testing agent workflows with this? Open an issue or join the Discord community with the prompt, workflow, or edge case you want supported.