arXiv MCP Server is an MCP-compatible service that enables natural-language interactions with the arXiv API. It offers tools to retrieve article metadata, download PDFs, search the arXiv catalog, and load articles into a large language model (LLM) context. Available tools include get_article_url (returns the URL for an arXiv article given its title), download_article (saves the article as a PDF on the local machine), load_article_to_context (injects article content into an LLM's context), get_details (fetches metadata for an article by title), and search_arxiv (performs keyword-based searches with fields such as all_fields, title, author, abstract, and start for pagination). The README also describes cross-platform setup via MacOS and Windows using the uv package manager, virtual environment creation with Python 3.13, and dependency synchronization. It shows integration with Claude Desktop via a configuration file (claude_desktop_config.json) and provides example prompts illustrating common use cases like retrieving details, listing papers, downloading PDFs, and performing searches. The server is intended to be run with uv, pointing to server.py, and supports a configurable DOWNLOAD_PATH.