Verified

Arxiv MCP Server

Search arXiv, retrieve LaTeX, and follow citation graphs for your AI research workflow.

Data last scanned yesterday · Reviewed last month

5811 tools★ 3,007PythonApache-2.0

Overview

Arxiv MCP Server provides tools to query and interact with the arXiv academic repository. It enables AI agents to search for papers, download documents, and extract specific LaTeX sections for deeper analysis. Users can also leverage the server to follow citation graphs and establish research alerts directly within their development environment. The server operates locally via stdio and relies on the uv package manager for installation and execution. By keeping index and local paper storage on the user machine, it facilitates a streamlined research process without the need for manual repository cloning or complex environment configuration.

Our verdict

With a security score of 55/100, this server is classified as high risk. The primary concern stems from the fact that it is a community-maintained project rather than an official vendor release, and critical aspects such as read-only mode, OAuth, and authentication mechanisms have not yet been formally reviewed for security. Due to these gaps in verification, it should be used with caution in environments where data isolation is a priority.

Supports zero direct dependencies to minimize supply chain surfaces
Active development with commits as recent as two days ago
Strong community engagement with 3007 GitHub stars
Large contributor base of 16 individuals reduces bus-factor risk

Lacks verification for authentication and OAuth support
Read-only mode status has not been officially reviewed
No verified homepage link provided in repository metadata

Tools

Tool	Description	Risk
search_papers	Search arXiv with optional category, date, and boolean filters.	low
download_paper	Download a paper by its arXiv ID.	medium
list_papers	List all papers downloaded locally.	low
read_paper	Read the full text of a locally downloaded paper in markdown.	low
get_paper_latex	Retrieve author-submitted LaTeX when available.	low
list_paper_latex_sections	Inspect a bounded page of the compact section outline.	low
get_paper_latex_section	Read one section by returned ID or exact title.	low
semantic_search	Semantic similarity search over your locally downloaded papers only.	low
citation_graph	Fetch references and citing papers via Semantic Scholar.	low
watch_topic	Save topic watches for newly published papers.	medium

Compatibility

Client	Local	Docker	Remote	Read-only
ChatGPT
Claude Desktop
Cursor
VS Code
Windsurf

Frequently asked questions

›What are the system requirements for running the ArXiv MCP Server?

The server requires Python 3.11 or higher to be installed on your system.

›Where does the server store downloaded papers?

Downloaded papers are saved locally, with the default storage location being ~/.arxiv-mcp-server/papers.

›What is the recommended security practice regarding prompt injection?

Because arXiv paper content is untrusted external input, you should use read-only MCP configurations and treat AI-generated summaries as data rather than instructions.

›Are there specific risks when using this server in multi-tool setups?

Yes, combining this server with filesystem, shell, or browser tools increases risk, as a prompt injection within a paper could potentially chain unintended tool calls.