openags/paper-search-mcp

What it solves

paper-search-mcp provides a unified interface for searching and downloading academic papers from a vast array of public and open-access sources. It solves the fragmentation of academic search, allowing users to query multiple platforms simultaneously and retrieve full-text PDFs through a prioritized fallback chain of open-access repositories.

How it works

The project uses a Two-Layer Architecture:

1. Unified Tooling: A high-level layer that handles concurrent multi-source searches, deduplication of results, and a download_with_fallback mechanism that tries various open-access paths before optional restricted sources.
2. Platform Connectors: Modular connectors for specific platforms (e.g., arXiv, PubMed, Semantic Scholar, OpenAlex). These connectors handle the specifics of each API or website, including DOI extraction and metadata retrieval.

It is implemented as a Model Context Protocol (MCP) server, meaning it can be integrated directly into LLM clients like Claude Desktop or used as a Claude Code skill via a CLI interface.

Who it’s for

Researchers, academics, and AI developers who need to automate the discovery and retrieval of academic literature for AI-driven research workflows.

Highlights

• Multi-Source Support: Integrates with over 20 sources including arXiv, PubMed, Google Scholar, Semantic Scholar, and Zenodo.
• Free-First Strategy: Prioritizes open and public data sources to minimize reliance on paid API keys.
• OA-First Fallback Chain: Automatically attempts to find legal open-access PDFs via sources like Unpaywall, CORE, and OpenAIRE before other options.
• MCP Integration: Seamlessly enhances LLM context by allowing AI agents to search and read papers directly.
• Extensible Design: New academic platforms can be added easily by extending the academic_platforms module.