ETF Flow MCP

The ETF Flow MCP fills a niche that many AI‑powered financial assistants lack: real‑time, granular insight into the buying and selling activity of cryptocurrency exchange‑traded funds (ETFs). By exposing a single, well‑structured tool——the server allows Claude and other LLMs to query historical flow data for Bitcoin (BTC) or Ethereum (ETH) ETFs with a simple, natural‑language prompt. This capability is essential for traders and analysts who need to gauge market sentiment, anticipate price movements, or build algorithmic strategies based on the net inflow and outflow of institutional capital.

At its core, the server connects to the CoinGlass API using a user‑supplied key and retrieves flow figures for each ETF ticker. The data is then reshaped into a Markdown pivot table: dates form the rows, each ETF ticker becomes a column, and an additional “Total” column aggregates all flows for that day. The table format is deliberately chosen because it is both human‑readable and easily parsed by downstream tools or scripts, enabling developers to embed the output directly into dashboards, reports, or further LLM prompts.

Key features of the MCP include:

• Unified Tool Interface – a single command that handles both BTC and ETH requests, simplifying client integration.
• Dynamic Data Retrieval – pulls the most recent historical data on demand, ensuring that AI agents always work with up‑to‑date market signals.
• Markdown Table Output – delivers results in a clean, tabular format that can be displayed directly in chat or exported for analytics.
• Prompt Guidance – the bundled helps craft user queries, reducing friction for non‑technical users.

Real‑world scenarios that benefit from this server abound. A portfolio manager can ask an AI assistant to “Show me the latest BTC ETF flow data in a table,” instantly receiving a snapshot that informs rebalancing decisions. A quantitative researcher might request “Get the ETH ETF flow history” to feed a time‑series model. Because the data is structured and machine‑friendly, developers can pipe it into Python notebooks, Power BI reports, or automated trading bots with minimal effort.

Integration is straightforward: the MCP registers itself as a tool in Claude Desktop’s configuration, exposing a hammer icon that users can click to trigger the command. Once connected, any LLM interaction that references ETF flow automatically routes through the MCP, ensuring consistent latency and reliability. The server’s lightweight Python implementation (Python 3.10+), combined with the fast uv package manager, makes deployment quick and maintainable.

In summary, the ETF Flow MCP turns abstract market sentiment into actionable data for AI assistants. By providing a clean, unified interface to ETF flow statistics and delivering the results in an immediately usable Markdown table, it empowers developers and traders alike to embed sophisticated financial insights into their AI workflows with minimal overhead.