Paper Search MCP
An MCP server for searching and downloading academic papers from arXiv, PubMed, bioRxiv, Semantic Scholar and other open sources
- 3,076 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Reaches external scientific services and downloads files to disk
- Optional API keys are sent to third-party servers
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/paper-search-mcpDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Reaches external scientific services and downloads files to disk
- Optional API keys are sent to third-party servers
This is third-party code. Review the repository files before installing.
What it does
The server gives an agent unified search and download of academic papers across many sources. A top layer combines concurrent multi-source search with deduplication and a download chain that runs from the source to open access and DOI resolution. A lower layer holds modular connectors for arXiv, PubMed, bioRxiv, medRxiv, Semantic Scholar, Crossref, OpenAlex, PubMed Central, Europe PMC, CORE, Zenodo, HAL and more. Results are normalized to one format, and for some papers PDF download and text extraction are available. The project follows a free-first principle: open databases come first, and API keys are optional, only widening limits and coverage. Beyond the MCP server, a Claude Code skill and a CLI over the same library are available.
Who it is for. For researchers, analysts and technical writers who gather academic literature through an agent.
Good fit when
- You need papers on a topic across several databases at once
- You need to download a PDF or extract text by identifier or DOI
- You need a source list that prioritizes open access
Not a fit when
- You need full text from paywalled databases you cannot access
- A plain web search is enough without installing a server
Example request
Find recent papers on CRISPR base editing in arXiv and Semantic ScholarMCP
- Transport
- stdio
- Authentication
- key is optional
Environment variables
PAPER_SEARCH_MCP_UNPAYWALL_EMAILrequired- Any valid email; without it the Unpaywall source is disabled.
PAPER_SEARCH_MCP_CORE_API_KEYsecret- Free CORE key that improves repository download stability.
PAPER_SEARCH_MCP_SEMANTIC_SCHOLAR_API_KEYsecret- Free Semantic Scholar key that raises request limits.
PAPER_SEARCH_MCP_GOOGLE_SCHOLAR_PROXY_URL- Proxy URL to bypass Google Scholar bot-detection.
PAPER_SEARCH_MCP_IEEE_API_KEYsecret- Activates the IEEE Xplore connector, otherwise a skeleton.
PAPER_SEARCH_MCP_ACM_API_KEYsecret- Activates the ACM Digital Library connector, otherwise a skeleton.
Limitations
All API keys are optional except Unpaywall, which needs an email in an environment variable or the source is disabled. Some sources are limited on the provider side: Google Scholar and SSRN block bots, and Semantic Scholar and CORE can hit rate limits without a key. The IEEE and ACM connectors are skeletons for now and activate only with their own keys. PDF download is mostly available for open access. The optional Sci-Hub connector is unstable, and responsibility for its use lies with the user.
How to disable. Remove the server from the client's MCP configuration. If you installed the package, remove it via pip or uv, and delete the Claude Code skill from the skills directory.
FAQ
Are API keys required?
No, except the Unpaywall email. Open databases work without keys, and keys for Semantic Scholar, CORE and others only raise limits and stability.
Can I install it as a skill instead of a server?
Yes. For Claude Code there is a separate skill over the paper-search CLI that works without MCP configuration.
How is Paper Search MCP different from searching a single source?
The server gives one parallel search across arXiv, PubMed, bioRxiv, Semantic Scholar and other open databases with result deduplication, plus a download chain for papers from available sources.
How do I connect an MCP for academic paper search?
Add the paper-search-mcp server with the uvx command to the client config, as in the install block. It works in Claude Desktop, Claude Code, Cursor and Codex. Some sources may need an API key, see the README.
Discussion
No questions or reviews yet.
Related
MCP serversPubMed MCP searches biomedical papers, retrieves available full text and formats citations with retraction and correction notices.
An MCP server and skill for Sciverse paper search: metadata, semantic retrieval, citations and full text for agents.
Medical MCP
medical-mcp
MCP server for medical sources: FDA and EMA drug data, WHO statistics, PubMed and clinical trials without API keys.
An MCP server for OpenAlex: 31 tools for papers, citations, authors and journals, plus a skill with a small CLI for Claude Code.
Connects a Zotero library to Claude and other agents: search papers, annotate, cite, and search by meaning
An MCP server gives agents search and answers over local documents using hybrid retrieval, reranking and section links.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Paper Search MCP (this card) | no | key is optional | 3,076 | |
| Google Scholar Skills | no | 545 | six skills and a gs-researcher agent | |
| Academic Paper Search | no | 388 | paper search across five databases with deduplication | |
| PaperGraph MCP | no | not required | 278 | an MCP server that splits math papers from arXiv |
README in short
The README describes a Python MCP server for searching and downloading academic papers from more than twenty sources with an open-access priority. The architecture has two layers: unified search with deduplication and a download chain on top, and modular per-database connectors below. It includes a capability matrix per source and a list of optional environment variables with keys. Installation is possible via Smithery, uvx, uv, pip, npx, Docker or from source, and for Claude Code there is a skill variant. MIT licensed.