Arxiv Mcp Server

Connects Claude directly to the arXiv preprint repository through four tools: free-text search with boolean operators and field prefixes, batch metadata fetching by paper ID, full-text HTML extraction with LaTeX math rendering, and category taxonomy browsing. Built on the author's mcp-ts-core framework with request queuing that respects arXiv's 3-second crawl delay and adaptive rate limit handling. Includes an optional local OAI-PMH mirror mode that caches metadata in SQLite to eliminate rate limit exposure for search and metadata operations. Available as a public hosted instance at arxiv.caseyjhand.com or self-hosted via stdio. Useful for literature reviews, citation gathering, or any workflow where you need to query and read academic papers without leaving your LLM conversation.