arXiv MCP Server Guide 2026: Give Claude Access to Research Papers

Table of Contents

1. What Is an arXiv MCP Server?
2. Why Give Claude Access to arXiv?
3. The Best arXiv MCP Servers
4. Comparison Table
5. Setup: Adding arXiv MCP to Claude
6. Research Workflows That Actually Work
7. What arXiv MCPs Can and Can't Do
8. Frequently Asked Questions

What Is an arXiv MCP Server?

An arXiv MCP server connects an AI client to arXiv — the open-access preprint repository that hosts millions of papers across physics, mathematics, computer science, quantitative biology, statistics, and more. Over the Model Context Protocol, the server exposes tools like search_papers, get_paper, download_paper, and read_paper, so Claude can find and analyze research inside a normal conversation.

The best part: arXiv's API is free and requires no authentication. Unlike most MCPs that wrap a paid or key-gated service, an arXiv MCP works the moment you install it. That makes it one of the friendliest entry points into the 75,000+ MCP servers indexed on Skiln, and a favorite of researchers, grad students, and anyone who reads papers to keep up.

Why Give Claude Access to arXiv?

Reading research is slow. An arXiv MCP does not replace careful reading, but it removes the friction around it:

• Find the right papers fast. Search by topic, author, or category and get back the most relevant recent preprints instead of scrolling listing pages.
• Read full text, not just abstracts. The leading servers download the PDF, convert it to text, and let Claude analyze the methodology and results in depth.
• Compare papers side by side. Pull three papers on the same problem and ask Claude to contrast their approaches and findings.
• Stay current. Ask for this week's new submissions in a category and get a digest of what's worth reading.

If your research also spans the open web and documentation, arXiv pairs naturally with other tools. Our research and academic MCP roundup shows how arXiv fits alongside web search, and the Fetch MCP guide covers pulling arbitrary URLs into context.

The Best arXiv MCP Servers

1. arxiv-mcp-server (blazickjp)

The most established arXiv MCP and the one to install first. It provides a full bridge between AI assistants and arXiv: search with filters, download papers, read the converted full text, and use built-in prompt templates for structured research tasks like deep-dive analysis. Written in Python and distributed for easy install via uvx.

Best for: Serious literature review, reading full papers end to end, structured research workflows.

Install: Register via uvx — see the setup section below.

2. arXiv-mcp (shoumikdc)

A clean, focused Model Context Protocol server for fetching the latest arXiv papers and querying them through an MCP client. Distributed via the Smithery platform, which makes one-line installation especially simple. A good pick if you want something lightweight and quick to set up.

Best for: Fast setup, staying on top of new submissions, lightweight querying.

Install: Install through Smithery, or clone and run directly.

3. arxiv-mcp (Tejas242)

A streamlined server that connects AI assistants to arXiv with a clear focus on three actions: search, analyze, and download research papers directly from your AI workflow. Minimal configuration, easy to read source, good for developers who want to understand exactly what the MCP does.

Best for: Developers who prefer a small, readable codebase and the core search-analyze-download loop.

Install: Clone the repo and register the server with your client.

4. Scientific Papers MCP (arXiv + OpenAlex)

Rather than arXiv alone, benedict2310/Scientific-Papers-MCP gives Claude real-time access to scientific papers from both arXiv and OpenAlex. OpenAlex adds a much broader index of published, peer-reviewed work with citation graphs, so this is the pick when preprints alone aren't enough.

Best for: Cross-source research, citation-aware literature reviews, fields where peer-reviewed coverage matters.

Install: Follow the repo README; no arXiv key needed, OpenAlex is also free.

Comparison Table

Setup: Adding arXiv MCP to Claude

Most arXiv MCPs are Python-based and install with uvx. For Claude Desktop, add to your config:

{   "mcpServers": {     "arxiv": {       "command": "uvx",       "args": ["arxiv-mcp-server"]     }   } }

Restart Claude Desktop and the arXiv tools appear. Because there's no API key, that's the entire setup.

For Claude Code:

claude mcp add arxiv -- uvx arxiv-mcp-server

Some servers accept a storage path so downloaded papers persist between sessions — check the README and pass it as an argument if you want a local paper cache. The Skiln Config Generator outputs ready-to-paste configs combining arXiv with your other research MCPs.

Research Workflows That Actually Work

• Topic scan. "Search arXiv for the most cited recent papers on retrieval-augmented generation and give me a one-paragraph summary of each." A reading list in seconds.
• Full-paper deep dive. "Download and read arXiv:2503.23278 and explain the threat taxonomy in plain language." Claude reads the whole PDF, not just the abstract.
• Comparison brief. "Pull three recent papers on MCP security and contrast their methodologies and conclusions." Useful for spotting where the field agrees and disagrees.
• Weekly digest. "What new cs.AI papers were submitted this week that touch agent evaluation?" Turn arXiv's firehose into a focused feed.
• Grounded writing. Combine arXiv with a web-search MCP like Perplexity or documentation via DeepWiki so Claude cites primary sources instead of guessing.

What arXiv MCPs Can and Can't Do

Set expectations correctly:

• arXiv is preprints. Many papers are excellent, but they are not all peer-reviewed. Treat findings accordingly, and use a cross-source server like Scientific-Papers-MCP when peer-review status matters.
• Coverage is field-dependent. arXiv is strongest in physics, math, CS, and quantitative fields. For medicine, law, or the humanities, you'll want other sources.
• Rate limits are informal but real. arXiv asks clients to be gentle with bulk downloads. Don't point an agent at "download every paper in this category" — it's inconsiderate to shared infrastructure and slow.
• PDF conversion is imperfect. Equations, tables, and figures may not survive the text conversion cleanly. For heavy math, keep the original PDF handy.

Frequently Asked Questions

What is an arXiv MCP server?

An arXiv MCP server is a Model Context Protocol server that connects Claude (or another MCP client) to arXiv, the open-access preprint repository for physics, math, computer science, and related fields. It exposes tools to search papers, fetch abstracts and metadata, and in most cases download and read the full PDF text, so Claude can analyze research directly inside a conversation.

Does the arXiv MCP need an API key?

No. arXiv provides a free, public API that requires no authentication, so arXiv MCP servers work with zero API keys. You only need the runtime the server is written in — usually Python via uvx or pip, or Node via npx. This makes arXiv one of the easiest MCPs to install.

Which arXiv MCP server is best?

blazickjp/arxiv-mcp-server is the most established and full-featured — it supports searching, downloading, reading, and analyzing papers, plus prompt templates for structured research. shoumikdc/arXiv-mcp is a clean option distributed via Smithery. Tejas242/arxiv-mcp is a streamlined choice focused on search, analyze, and download. For cross-source research, Scientific-Papers-MCP adds OpenAlex alongside arXiv.

Can Claude read the full text of an arXiv paper?

Yes, with most servers. The leading arXiv MCPs download the paper PDF, convert it to text (often markdown), and store it locally so Claude can read the full content, not just the abstract. This lets you ask questions about the methodology, summarize the results, or compare several papers section by section.

Is the arXiv MCP free?

Yes. Every arXiv MCP server is free and open source, and arXiv's own API is free and unauthenticated. There is no cost beyond your normal Claude usage. Be considerate with request volume, since arXiv asks clients to rate-limit bulk downloads out of respect for shared infrastructure.

Does arXiv MCP work with Claude Code and Cursor?

Yes. The Model Context Protocol is a shared standard, so an arXiv MCP server works in Claude Desktop, Claude Code, Cursor, Windsurf, Cline, and Zed. Python-based servers install with uvx or pip; you register them with claude mcp add for Claude Code or a JSON entry for Claude Desktop.

Can I use arXiv MCP for a literature review?

Absolutely — it's one of the best use cases. Ask Claude to search a topic, pull the ten most relevant recent papers, download and read each one, and produce a synthesized summary with citations. Pair it with a note-taking MCP to save the output. It compresses hours of reading into a structured brief you can verify against the sources.

Where can I find other research MCP servers?

Skiln indexes every research and academic MCP across all major registries. See our best research and academic MCP servers roundup to compare arXiv against web-search tools like Perplexity, Exa, and Tavily, plus documentation tools like DeepWiki and Fetch.

Last updated: July 02, 2026 · Skiln tracks MCP server releases daily across 13 source registries.