openags/paper-search-mcp

What it solves

paper-search-mcp simplifies the process of finding and retrieving academic papers across a fragmented landscape of repositories. It eliminates the need to manually search multiple databases by providing a unified interface to search, download PDFs, and extract text from dozens of academic sources, prioritizing free and open-access data.

How it works

The project uses a two-layer architecture to handle academic data:

1. Unified Tooling Layer: Provides high-level functions like search_papers for concurrent searching and deduplication across multiple sources, and download_with_fallback which attempts to retrieve a PDF through a sequence of open-access links and repositories.
2. Platform Connectors Layer: Modular connectors for specific platforms (such as arXiv, PubMed, Semantic Scholar, and OpenAlex) that handle the specifics of each API or website, including DOI extraction and metadata standardization.

It can be deployed as a Model Context Protocol (MCP) server for LLM clients (like Claude Desktop) or used as a CLI tool. It supports various transport methods including stdio, sse, and streamable-http.

Who it’s for

• Researchers who need a centralized way to discover and download academic literature.
• AI Developers building LLM-driven research workflows that require grounded academic context.
• Users of MCP-compatible clients who want to give their AI assistants the ability to search and read academic papers.

Highlights

• Extensive Source Support: Integrates with arXiv, PubMed, bioRxiv, medRxiv, Google Scholar, Semantic Scholar, Crossref, OpenAlex, and many others.
• Free-First Strategy: Prioritizes public and open data sources over paid or restricted ones.
• Open-Access Fallback Chain: Automatically tries multiple open-access paths (e.g., Unpaywall, CORE, OpenAIRE) to find a legal PDF before optional fallbacks.
• LLM-Ready: Standardizes output into a consistent Paper class for easy consumption by AI models.
• PDF Section Extraction: Includes a local utility to split downloaded PDFs into recognized standalone headings for better text analysis.