Adamamer20 Paper Search MCP Openai

What adamamer20-paper-search-mcp-openai does

The adamamer20-paper-search-mcp-openai MCP server exposes academic literature search and download functions through the Model Context Protocol. It covers arXiv, PubMed, bioRxiv, medRxiv, Google Scholar, the IACR ePrint Archive, and Semantic Scholar. The project also describes optional Sci-Hub support, while several other platforms remain listed as planned rather than available.

Search results use a consistent dictionary-shaped representation through the project’s Paper class. This gives an MCP client a common structure for handling results from different providers instead of requiring each source to use a separate response format.

How it works

The server is written in Python and built with the MCP Python SDK. It performs network requests asynchronously with httpx, which lets its tools communicate with external academic services without blocking the server’s request handling.

Provider-specific operations include tools such as search_arxiv and download_arxiv. The project also provides standardized search and fetch tools intended for OpenAI Deep Research and ChatGPT connectors. The exact source selected depends on the search or retrieval operation requested by the client.

The adamamer20-paper-search-mcp-openai MCP server is designed to run as an MCP process launched by a compatible client. The supplied Claude Desktop configuration starts it with uv, points to the installation directory, and invokes the paper_search_mcp.server module.

Setup and configuration

Install the package with uv add paper-search-mcp. The documented Claude Desktop setup adds an MCP server entry to the platform’s configuration file, using uv run, the project directory, and the server module as the launch arguments. Replace the example installation path with the local path where the package or checkout resides.

A development setup uses uv, a virtual environment, and an editable installation. The repository instructions also show optional installation of development dependencies such as pytest and flake8.

SEMANTIC_SCHOLAR_API_KEY can be supplied in the client environment for enhanced Semantic Scholar features, but the configuration marks it as optional. The project is licensed under MIT and requires Python 3.10 or newer.

Tools and capabilities

• Search academic sources including arXiv, PubMed, bioRxiv, medRxiv, Google Scholar, IACR ePrint Archive, and Semantic Scholar.
• Download papers, including provider-specific PDF retrieval such as download_arxiv.
• Use common search and fetch operations for Deep Research and ChatGPT connector workflows.
• Return paper data in a consistent format through the Paper class.
• Extend source coverage by adding platforms under the project’s academic-platform modules.

Limitations and notes

The available source list should be distinguished from the planned integrations. PubMed Central, ScienceDirect, Springer Link, IEEE Xplore, ACM Digital Library, Web of Science, Scopus, JSTOR, ResearchGate, CORE, and Microsoft Academic are listed as future work in the repository material.

Semantic Scholar’s enhanced features may require an API key. The documentation specifically demonstrates Claude Desktop configuration; other MCP clients may require their own equivalent server configuration. The adamamer20-paper-search-mcp-openai MCP server depends on network access to the academic sources it queries and downloads from.