arxiv-latex
MCP server that uses arxiv-to-prompt to fetch and process arXiv LaTeX sources for precise interpretation of mathematical expressions in scientific papers.
claude mcp add --transport stdio takashiishida-arxiv-latex-mcp uv --directory /ABSOLUTE/PATH/TO/arxiv-latex-mcp run server/main.py
How to use
This MCP server provides access to arXiv LaTeX sources through MCP clients such as Claude Desktop, Claude Code, Cursor, and others. It leverages the arxiv-to-prompt project under the hood to fetch and process LaTeX sources from arXiv, enabling you to query and extract structured information from papers, including full LaTeX content, abstracts, and individual sections. The available MCP tools exposed to Claude Desktop include get_paper_prompt (full flattened LaTeX for a paper), get_paper_abstract (the abstract only), list_paper_sections (section headings), and get_paper_section (a specific section by path). This setup is particularly beneficial for accurately interpreting mathematical formulas and notations directly from the source rather than from rendered PDFs.
To use it, ensure your MCP client is configured to connect to the arxiv-latex MCP server using the provided command and path. Once connected, select one of the listed MCP tools in your client’s Available MCP tools list and provide the target paper identifier (e.g., arXiv ID) and any required parameters. The tools will return the requested LaTeX content or metadata, which you can then feed into your prompts or workflows for analysis, summarization, or explanation.
How to install
Prerequisites:
- Python 3.8+ installed on your system
- Access to install or run uv (uVicorn/uvicorn-like) according to your platform
- An arXiv paper you want to fetch via LaTeX sources
Step-by-step installation:
- Install uv and necessary dependencies (adjust for your OS):
# Example: using pipx to install uv if available, or install via your package manager
# If uv is a Python package, you can install as a module runner
pipx install uv
- Clone or download the arxiv-latex MCP repository to your local machine where you will run the server.
- Ensure you have a valid absolute path to the MCP server directory. The configuration expects the following entry (as shown in the README):
{
"mcpServers": {
"arxiv-latex-mcp": {
"command": "uv",
"args": [
"--directory",
"/ABSOLUTE/PATH/TO/arxiv-latex-mcp",
"run",
"server/main.py"
]
}
}
}
- Update the path in the config to point to your actual directory, then save the config.
- Start your MCP hub/application with the updated configuration and verify the arxiv-latex-mcp service appears in the list of available MCP tools in Claude Desktop or your client of choice.
Note: If uv is not found, locate the full path to uv on your system and replace the command field accordingly, using which uv (Mac/Linux) or where uv (Windows) to determine the path.
Additional notes
Tips and common issues:
- If you see connection errors, verify that the path provided in --directory is accessible and that server/main.py is the correct entry point for starting the MCP server locally.
- Ensure Python dependencies required by server/main.py are installed in your environment (e.g., via a virtual environment).
- For Claude Desktop users, the listed tools (get_paper_prompt, get_paper_abstract, list_paper_sections, get_paper_section) will appear after the MCP server is connected; ensure you have the permissions to access arXiv content and that any network proxies are correctly configured.
- If the arXiv paper ID format changes, ensure your client passes the correct identifier and that the arxiv-to-prompt integration remains compatible with the current arXiv API.
- Replace the placeholder path /ABSOLUTE/PATH/TO/arxiv-latex-mcp with the actual absolute path to the repository on your machine.
- Restart the application after saving the configuration to ensure the MCP server is loaded properly.
Related MCP Servers
mcp-aktools
📈 提供股票、加密货币的数据查询和分析功能MCP服务器
packt-netops-ai-workshop
🔧 Build Intelligent Networks with AI
lmstudio-toolpack
A MCP stdio toolpack for local LLMs
txtai-assistant
Model Context Protocol (MCP) server implementation for semantic vector search and memory management using TxtAI. This server provides a robust API for storing, retrieving, and managing text-based memories with semantic vector database search capabilities. You can use Claude and Cline AI as well.
argus
AI code review MCP server with Zero-Trust approach. Multi-model support, caching, multilingual. Works with Windsurf & Claude Desktop
openapi -proxy
An MCP server that provides tools for exploring large OpenAPI schemas