paper-finder-student-mcp

Paper Finder MCP

Google Scholar for your AI. See how it compares.

Paper Finder MCP is an academic paper search MCP server. Ask your AI to find real scholarly papers for an essay or research project, with authors, year, citation counts, DOIs, a one line abstract and a free to read link when one exists.

Sort by citation count to get the most cited papers on a topic, or by date to get the newest ones. Ask for the papers a paper cites, or the papers that cite it, when you build a literature review.

A search that matches thousands of papers says how many there are and which results you are reading, so you can keep going with the next page instead of stopping at the first handful. The author, citing, related and reference lists work the same way.

Results come straight from OpenAlex, a free open catalog of scholarly works, so the papers are real, not invented. Free forever, no account and no key.

Students often type things like 'find 5 sources for my essay on spaced repetition' or 'citations for my literature review'.

Install

Click a button to add the server to Cursor or VS Code, or set it up by hand with the config below.

npx -y paper-finder-student-mcp

Add this to your AI app's MCP config. Where the file goes and what the key inside it is called depends on the app.

Claude Desktop

The file is claude_desktop_config.json.

{
  "mcpServers": {
    "paper-finder": {
      "command": "npx",
      "args": ["-y", "paper-finder-student-mcp"]
    }
  }
}

Cursor

The file is ~/.cursor/mcp.json for every project, or .cursor/mcp.json for one project.

{
  "mcpServers": {
    "paper-finder": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "paper-finder-student-mcp"]
    }
  }
}

VS Code

The file is .vscode/mcp.json in your workspace. VS Code calls the key servers and asks for the type.

{
  "servers": {
    "paper-finder": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "paper-finder-student-mcp"]
    }
  }
}

Restart your AI app, then try one of the prompts below. You need Node.js 20 or newer. For Claude Code, run claude mcp add paper-finder -- npx -y paper-finder-student-mcp.

What it looks like

This is what search_papers returns for the query "spaced repetition" when you ask for the most cited papers first.

Papers about "spaced repetition", most cited first
Showing 1 to 3 of 12,408. Pass page 2 for the next page.

1. Spaced Repetition Promotes Efficient and Effective Learning (2016)
  Sean H. K. Kang, Policy Insights from the Behavioral and Brain Sciences
  405 citations. DOI 10.1177/2372732215624708. W2310690298
  No free copy found
  Abstract: Concern that students in the United States are less proficient in mathematics, science, and reading than their peers in other countries has led some to question whether American students spend enough...

2. A Trainable Spaced Repetition Model for Language Learning (2016)
  Burr Settles, Brendan Meeder, Annual Meeting of the Association for Computational Linguistics (ACL)
  217 citations. DOI 10.18653/v1/p16-1174. W2514897959
  Free to read, https://www.aclweb.org/anthology/P16-1174.pdf
  Abstract: We present half-life regression (HLR), a novel model for spaced repetition practice with applications to second language acquisition.HLR combines psycholinguistic theory with modern machine learning...

3. Enhancing human learning via spaced repetition optimization (2019)
  Behzad Tabibian, Utkarsh Upadhyay, Abir De, et al., Proceedings of the National Academy of Sciences
  181 citations. DOI 10.1073/pnas.1815156116. W2914319828
  Free to read, https://doi.org/10.1073/pnas.1815156116
  Abstract: Spaced repetition is a technique for efficient memorization which uses repeated review of content following a schedule determined by a spaced repetition algorithm to improve long-term retention.

Each result carries the title, the authors, the year, the venue, the citation count, the DOI, the OpenAlex id, a free to read link when one exists and the first line of the abstract, so you can judge a paper before you open it. The showing line says which results you are reading and how many matched in all, and the page to ask for next. OpenAlex counts change over time, so the totals in this sample will differ when you run it.

Example prompts

• "Find open access papers on spaced repetition and exam performance since 2020."
• "Find the most cited papers on spaced repetition since 2020."
• "Find papers with spaced repetition in the title."
• "Find papers on spaced repetition published from 2015 to 2020 with at least 50 citations."
• "Show me the newest journal articles about spaced repetition."
• "Find 5 sources for my essay on spaced repetition."
• "Citations for my literature review."
• "Get the paper with DOI 10.1038/nature14539 and show me the abstract."
• "Find a free PDF of this paper, DOI 10.1038/nature14539."
• "What papers are related to DOI 10.48550/arXiv.1706.03762?"
• "Who cited DOI 10.48550/arXiv.1706.03762?"
• "What does DOI 10.1038/nature14539 cite?"
• "List papers by John Dewey on education since 1950."
• "List papers by A. Duckworth on grit."

How it compares

Three ways to get papers for an essay, and what each one actually gives you. This section has a stable anchor at README.md#how-it-compares if you want to link someone here.

	Paper Finder MCP	Asking ChatGPT	Semantic Scholar MCP
Where the answers come from	OpenAlex, queried live each time you ask	Whatever the chat app does on its own, with no fixed catalog behind the answer	The Semantic Scholar catalog, queried live each time you ask
Can you open the source	Yes, every result carries a DOI or OpenAlex id you can open	A paper it names may not exist, and there is often nothing to open	Yes, results carry identifiers you can open
Free to read copy	find_open_access_pdf returns a link when one exists, or says plainly that none was found	It may offer a link, but nothing checks it against a catalog	Depends on what that server exposes, see its own README
Citations and related work	find_citing_papers, paper_references, find_related_papers and list_author_papers, all in the same chat	You can ask, but you get one guess at a time, not a list read from an index	It has its own tools, check that project's README for which ones
Account, key or plan	None, no account and no key	You need an account with that service to ask	Depends on that project, check its own docs
Where it runs	Inside your MCP client, next to your other tools	In a chat window	Inside your MCP client

Paper Finder MCP is the one where every line has a live catalog record behind it and the follow up questions are tools rather than more guessing. The Semantic Scholar MCP column describes those servers in general, since they are separate projects with their own features and setup. Always open the paper itself before you cite it.

Tools

Tool	What it does	Parameters
search_papers	Search scholarly works by keywords, with optional year range, citation count, open access, title matches, type and order settings	query, from_year, to_year, min_citations, open_access_only, title_only, type, sort, page, limit
get_paper	Details for one paper from its DOI, doi.org link or OpenAlex id, including the abstract	paper, doi
find_open_access_pdf	The free to read copy of one paper, preferring a PDF link, or a plain note that none exists	paper
find_related_papers	Papers OpenAlex marks as related to a paper	paper, page, limit
find_citing_papers	Newer papers that cite a paper, most recent first	paper, page, limit
paper_references	The papers a paper cites, so you can read its bibliography	paper, sort, page, limit
list_author_papers	Papers by one author, most cited first, with the name looked up in the OpenAlex author records	author, from_year, open_access_only, page, limit

Every tool description states its inputs and one example prompt, so clients can pick the right tool without reading this page. paper accepts a DOI like 10.1038/nature14539, a doi.org link, or an OpenAlex id like W2741809807. doi accepts a DOI or doi.org link with no OpenAlex ids, and get_paper takes either paper or doi, so a DOI lookup has one home. limit is 1 to 25, default 8 and page is 1 to 400, default 1. search_papers shows the title, authors, year, venue, citations, DOI and a one line abstract for every hit, so you can judge a paper before you open it. sort takes relevance (the default OpenAlex order), cited_by_count:desc for the most cited papers on the topic first, or publication_date:desc for the newest papers first. type takes journal_article, preprint or book_chapter to keep the results to one kind of paper. title_only keeps only the papers whose title carries the search words, which stops a famous paper that only mentions the topic from taking the top spot on a broad search. from_year and to_year set the two ends of a range of years, so a search can ask for the papers published from 2015 to 2020, and min_citations keeps only the papers with at least that many citations, so a search can ask for the well cited papers on a topic. list_author_papers looks the name up in the OpenAlex author records first and then takes the papers from that author id, so J. Smith, Smith, John and John Smith all reach the same papers, even when the name on the paper is written differently. When the name cannot be resolved that way, the tool matches the name as it is written on the papers instead. find_citing_papers gives the papers that cite a paper, and paper_references gives the papers it cites, so a literature review can walk the citation graph in both directions. Those two lists and the related paper list show the same one line abstract under every entry, so a paper can be judged from the list before it is opened. paper_references takes its own sort, with cited_by_count:desc for the most cited references first and publication_year:asc for the oldest references first, which reads like a bibliography from the beginning. The heading says how many references OpenAlex holds for the paper, so a list cut short by limit is clear. Every search and every one of those lists comes with a showing line that says which results are on this page and how many matched in all, and names the page to ask for next, so a long list can be read a page at a time with page.

cited_by, related_papers, get_paper_by_doi and get_related_work still exist as deprecated aliases so saved prompts keep working. They do the same work as find_citing_papers, find_related_papers and get_paper with doi, and new prompts should use those instead. They are not in the tool list a client sees, so a client gets the tools above and no second name for them. A saved prompt that calls one of the old names still gets an answer.

Good to know

• Lookups that miss are flagged as errors with a stable code, not empty results. A wrong or unknown reference returns a not found error that suggests searching by the title or checking the DOI format, and a rate limit returns an error that says to wait about a minute and retry. An OpenAlex outage, such as a server error, a request that runs out of time or a connection that breaks, returns an upstream error with the code upstream_error and says to try again shortly, so a client can tell an outage from a bad request. A missing free copy is not an error, find_open_access_pdf just says that none was found.
• sort reorders the papers OpenAlex matched. It does not change which papers match, so on a broad topic the most cited match can be a famous paper that only mentions the topic. Ask for relevance when you want the closest match first, or set title_only to true so only the papers with the search words in the title come back.
• from_year, to_year and min_citations narrow the results, so a year range that is too narrow or a citation count that is too high can leave you with nothing. When a search finds nothing it says which of those filters to relax.
• Author names go through the OpenAlex author records, so a name that is written in another form on the papers still finds them. OpenAlex sometimes keeps more than one profile for a name. In that case the tool shows the profile whose name matches best and has the most papers, and it lists the others. You can pass one of those ids, such as A5108843829, as author to see that profile.
• Citation counts and open access links come from OpenAlex and can be a little behind the publisher.
• If you want faster, more reliable service for heavy use, set OPENALEX_EMAIL to your own email. OpenAlex asks for it so they can reach you, and it is sent only to OpenAlex.
• OPENALEX_BASE_URL points the server somewhere else, which the tests use.
• Always read the paper yourself before you cite it. A catalog entry is not the paper.

Development

npm install
npm run build
npm test

Tests run against a fake OpenAlex server, so they need no network.

More free student MCP servers

All free, open source and read only where they touch your accounts. Browse the full list at brightspacemcp.com/more-mcps.

Server	What it does
Brightspace MCP	D2L Brightspace for AI. Due dates with real submission status, grades, a target grade calculator and course files.
Canvas MCP	Canvas LMS for AI, read only. Due dates, missing work, grades and announcements.
Moodle MCP	Moodle for AI, read only. Due dates, overdue work, grades and announcements.
Google Classroom MCP	Google Classroom for AI, read only. Todos, due dates, missing work and grades.
GPA Calculator MCP	GPA, grade conversion and the grade you need on the final, for AI. Works offline, no account.
Citations MCP	APA, MLA and Chicago citations from a DOI, an ISBN or typed details. No account.
Study Planner MCP	Exam study schedules, spaced repetition dates, pomodoro plans and assignment milestones. Works offline.

License

MIT