Everyrow MCP Server

This connects Claude to FutureSearch's multi-agent research infrastructure. You get six operations: forecast() for predictions with dates and probabilities, multi_agent() to dispatch research teams on single questions, agent_map() to run one researcher per row across thousands of entries, plus rank(), classify(), and dedupe() for scoring and categorizing at scale. The practical use case is when you need to process large datasets where each row requires web research or judgment, like finding FDA status for 10,000 drugs or scoring 500 startups. It handles the orchestration of parallel agent execution and returns structured results back to Claude. Requires a FutureSearch API key and works through the standard MCP stdio transport.