# arxiv-mcp-server

A local MCP server for AI agents doing literature research on arXiv, reading original LaTeX sections one at a time, exporting BibTeX citations, and keeping topic watches.

arxiv-mcp-server is a local MCP server for agent literature work on arXiv. Its distinguishing features are reading a paper's original LaTeX section by section, exporting BibTeX citations from authoritative arXiv metadata, and keeping topic watches, with papers stored on disk and search treated as optional rather than the main feature; the intended working loop is paper ID, then outline, then one section, then citations.

It installs with uvx and needs no repository clone or separate Python environment, adding a small stdio configuration block to clients such as Claude Desktop, Claude Code, Codex, Kiro or VS Code. A packaged plugin for Claude Code additionally installs a bundled arXiv research skill alongside the MCP connection. The default paper storage directory can be overridden with a command-line argument.

It is published on PyPI, is listed on the official MCP registry, and is released under the Apache 2.0 license.

Website: https://www.pulsemcp.com/servers/blazickjp-arxiv-mcp-server