This repository implements an MCP (Model Context Protocol) server that connects AI agents to Exa's web search, content fetching, and multi-step research capabilities. It is deployed as a hosted HTTP endpoint on Vercel (Streamable HTTP transport) and also supports local stdio and standalone HTTP runtime modes. The server exposes a set of MCP tools including web_search_exa, web_fetch_exa, web_search_advanced_exa, and agent_run, each backed by the Exa API via the exa-js client. It handles authentication via API key or OAuth JWT, rate limiting via Upstash Redis, and analytics via Agnost. The codebase is well-structured with clean separation between tool registration, API client configuration, response sanitization, and error handling. It includes comprehensive unit tests and CI/CD pipelines for npm publishing and Docker image deployment. The architecture is a single server that can be deployed in multiple environments, with per-request configuration via URL parameters and headers.