MCP Servers · blazickjp

arxiv-mcp-server

arxiv-mcp-server is a Python bridge that lets AI assistants search and read arXiv research papers through the Model Context Protocol. It stores papers locally and integrates with Claude, ChatGPT, and other LLM clients via a standard MCP interface.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/blazickjp/arxiv-mcp-server

2.9k

GitHub stars

235

Forks

Python

Primary language

Apache-2.0

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	blazickjp/arxiv-mcp-server
Owner	blazickjp
Primary language	Python
License	Apache-2.0 — OSI-approved
Stars	2.9k
Forks	235
Open issues	16
Latest release	v0.5.0 (2026-04-26)
Last updated	2026-05-18
Source	https://github.com/blazickjp/arxiv-mcp-server

What arxiv-mcp-server is

An MCP server implementation (Python 3.11+, Apache 2.0) that wraps arXiv API calls, handles HTML/PDF paper retrieval, and exposes search/access/listing tools over stdio. Deployable via uv tool, Claude Desktop bundles (.mcpb), or manual MCP configuration.

Quickstart

Get the arxiv-mcp-server source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/blazickjp/arxiv-mcp-server.git2cd arxiv-mcp-server3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

AI-Assisted Literature Review

Researchers using Claude or GPT can search arXiv directly within the chat interface, fetch summaries, and ask the AI to compare papers—eliminating manual searches and copy-paste steps.

Research Team Knowledge Base

Teams can configure a shared MCP server pointing to a central paper storage directory, enabling consistent paper access across multiple AI clients and reducing duplication.

Agentic Pipelines with Controlled Paper Discovery

Build multi-tool workflows (e.g., search papers → summarize → trigger downstream analysis) where paper discovery is mediated through a read-only MCP connection to prevent injection attacks.

Implementation considerations

Use `uv tool install arxiv-mcp-server` (not npm or pip install in a venv) to ensure the executable is globally available and properly isolated.
For older arXiv papers (PDF-only), install the optional `[pdf]` extra: `uv tool install 'arxiv-mcp-server[pdf]'` to enable pymupdf4llm support.
Configure a dedicated `--storage-path` for downloaded papers to avoid filling user home directories and to enable sharing across team members.
Review the README's OWASP LLM01/AG01 mitigations: use read-only MCP configs where possible, review AI summaries before acting on them, and treat paper content as untrusted input.
Test the integration with your target MCP client (Claude Desktop, VS Code, Kiro, Codex) before deployment to verify stdio communication and tool availability.

When to avoid it — and what to weigh

High-Risk Multi-Tool Workflows Without Isolation — Do not use in agentic setups that combine unrestricted paper reading with shell, filesystem, or network access tools on the same client. Malicious paper content could chain tool calls.
Requirement for Real-Time Paper Modification — This server is read-only. If you need to annotate, edit, or push changes to arXiv papers, this is not the right tool.
Offline or Air-Gapped Deployments — The server requires live arXiv API access. It does not support pre-cached offline paper bundles or disconnected operation.
Highly Sensitive Proprietary Workflows — If papers will be fed into models you do not control (e.g., public ChatGPT), or if outputs require audit trails, review prompt injection mitigations and model endpoint policies first.

License & commercial use

Apache License 2.0: a permissive OSI-approved license allowing commercial use, modification, and distribution, provided that license notices and changes are documented. No strong copyleft obligations.

Apache 2.0 permits commercial use without royalties. You may use, modify, and integrate this server into proprietary products. Attribution is not required, but license and copyright notices must be retained in distributions. Requires legal review if bundling with proprietary LLM clients or reselling as part of a commercial service.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Strong
License clarity	Clear
Deployment complexity	Low
DEV.co fit	Good
Assessment confidence	High

Security considerations

README explicitly documents prompt injection risk: untrusted arXiv paper content is passed directly into model context, and a maliciously crafted paper could embed adversarial instructions to override behavior, exfiltrate data, or invoke other tools. OWASP LLM01/AG01 references included. Recommended mitigations: use read-only MCP configs, review AI summaries before acting on them, isolate from write-capable tools, and treat outputs as data not instructions. No other security advisories, CVEs, or authentication mechanisms detected. No end-to-end encryption or input sanitization in the scope.

Alternatives to consider

arXiv Python API + custom LLM integration

Gives you full control over paper fetching and model prompts, but requires building and maintaining your own MCP server or integration layer.

Semantic Scholar API with plugin/extension

Broader scope (cross-publisher metadata, citations, author info) but less direct integration with MCP clients and steeper API learning curve.

Private document ingestion (e.g., LlamaIndex, Langchain with local PDFs)

Avoids external API dependency and prompt injection risk if papers are pre-vetted, but requires manual paper collection and no native arXiv search.

Software development agency

Build on arxiv-mcp-server with DEV.co software developers

Install arxiv-mcp-server today and give your Claude or LLM client direct access to research papers. Start with one-click Claude Desktop setup or manual MCP config.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

mysql_mcp_server mcp-server-qdrant mcp-notion-server

Related on DEV.co

Explore the category and the services that help you build with it.

MCP Servers AI Application Development API Development Custom Software Development Talk to DEV.co

arxiv-mcp-server FAQ

Can I use this with OpenAI's ChatGPT or only Claude?

The server implements the MCP protocol, which Claude Desktop supports natively. ChatGPT does not yet support MCP servers directly; you would need a third-party integration layer or a custom wrapper. Confirm with your MCP client documentation.

How are papers stored, and can multiple team members share them?

Papers are stored as files in the local `--storage-path` directory. For team sharing, mount a network drive (NFS, SMB) or cloud storage (S3) at that path and ensure file permissions allow read access. Each user's MCP server instance reads from the same location.

What happens if a paper is malicious or contains adversarial content?

The README warns that paper content is untrusted. The server passes it directly to the model context; mitigations include: using read-only MCP configs to prevent the AI from invoking other tools, reviewing AI summaries critically before acting on them, and isolating this server from filesystem/shell access in multi-tool setups.

Is there an HTTP endpoint or can I call this from outside MCP?

Not from the provided data. The server is MCP-only, communicating via stdio. If you need HTTP access, you would need to build a wrapper or use an alternative arXiv integration.

Custom software development services

DEV.co is a software development agency delivering custom software development services to companies building on open source. Our software developers and web developers design, integrate, and ship production systems — spanning web development, APIs, AI, data, and cloud. If arxiv-mcp-server is part of your mcp servers roadmap, our team can implement, customize, migrate, and maintain it.

Enable Your AI Assistant to Search arXiv

Install arxiv-mcp-server today and give your Claude or LLM client direct access to research papers. Start with one-click Claude Desktop setup or manual MCP config.

Talk to DEV.co Schedule a Consultation