An MCP server that delivers crypto ETF flow data (BTC & ETH) in a ready-to-use table format for AI agents.
https://github.com/kukapay/etf-flow-mcpYou know the drill: your AI agent needs crypto ETF flow data, but you're stuck writing parsers, formatting tables, and dealing with inconsistent APIs. Meanwhile, you just want clean, structured data that works immediately.
The ETF Flow MCP server eliminates that busywork entirely. It pulls BTC and ETH ETF flow data from CoinGlass and delivers it as properly formatted markdown tables, ready for your AI agents to analyze without any preprocessing.
No More Data Formatting Hell: Instead of writing custom parsers for ETF data, you get clean pivot tables with ETF tickers as columns, dates as rows, and totals automatically calculated. The data arrives exactly how you'd want to present it.
Direct AI Agent Integration: Your Claude Desktop setup gets immediate access to current ETF flows through natural language queries like "Show me today's BTC ETF flows" or "Compare this week's ETH vs BTC ETF activity."
Production-Ready Output: The markdown tables are formatted for immediate use in reports, analysis, or feeding into other systems. No reformatting, no cleanup required.
Trading Signal Development: Pull ETF flow patterns to identify institutional sentiment shifts. Query historical flows to backtest strategies or spot anomalies in real-time.
Market Analysis Automation: Build workflows where your AI agents automatically analyze ETF flows alongside price action, generating insights about institutional money movements.
Research and Reporting: Generate daily/weekly ETF flow summaries for internal teams or clients without manual data gathering and table creation.
Alert Systems: Create conditional logic based on ETF flow thresholds – when IBIT sees massive inflows or ETHE shows significant outflows, your system can trigger notifications or analysis workflows.
The server integrates directly into Claude Desktop with a simple JSON config. Add your CoinGlass API key, restart Claude, and you're querying ETF data through natural language immediately.
For custom applications, the MCP protocol means your agents can call the get_etf_flow tool programmatically, specifying BTC or ETH and getting back structured data every time.
Clone the repo, run uv sync to handle dependencies, add your CoinGlass API key to the environment, and configure Claude Desktop. The whole setup takes under 5 minutes.
The server uses Python 3.10+ with minimal dependencies (pandas, requests) and integrates with uv for fast package management. It's designed to be lightweight and reliable for production use.
When your AI agent needs to understand crypto institutional flows, this MCP server delivers the data in exactly the format you need, when you need it. No data wrangling, no formatting delays – just clean ETF flow intelligence ready for immediate analysis.