Skip to content
Kitbay
MCP server

arxiv-mcp-server

Runs locallyApache-2.0Updated on Oct 8, 20263,271GitHub starsNo API key needed

A local MCP server by Joseph Blazick for literature work on arXiv, installed from PyPI and started with uvx.

What it does

A local MCP server by Joseph Blazick for literature work on arXiv, installed from PyPI and started with uvx. It can search arXiv, fetch abstracts, download papers and convert them to Markdown on disk, then read them back in bounded chunks or one section at a time. It can also pull the author's original LaTeX source section by section, export BibTeX from arXiv metadata, follow references and citing papers through the Semantic Scholar API, and keep saved topic watches that report new papers. Optional extras add PDF conversion and local semantic search over downloaded papers. It runs over stdio by default, with an optional Streamable HTTP mode bound to localhost.

Capabilities

  • Search arXiv by query, category and date range
  • Download a paper and read it in bounded chunks or by section
  • Read the original LaTeX source one section at a time
  • Export BibTeX for one or more arXiv IDs
  • List references and citing papers via Semantic Scholar
  • Save topic watches and check them for new papers