# Paper Search MCP

A Python MCP server, CLI and Claude Code skill for searching and downloading academic papers from sources such as arXiv, PubMed and bioRxiv, favoring free and open sources and returning results standardized for downstream use by a language model.

Paper Search MCP searches and downloads academic papers from multiple sources, and extracts text from the resulting PDFs, for use by researchers or in AI-driven research workflows. It runs as an MCP server for Claude Desktop and other MCP clients, or as a Claude Code skill with its own command-line interface.

It follows a free-first strategy, defaulting to open and public sources over paid or restricted ones, and only using an API key where one improves stability, rate limits or metadata quality; the MCP still works without keys wherever possible. A two-layer architecture separates a high-level, multi-source concurrent search tool from source-specific connectors, and results are deduplicated and standardized for downstream use by a language model. The bundled server currently runs locally over stdio and does not implement OAuth-based authorization on its own; a protected remote deployment needs a separate gateway or reverse proxy in front of it.

Website: https://github.com/openags/paper-search-mcp