ArXiv MCP Server

Researchstdiohttp-streamingv0.5.0
Repository

MCP server for searching and accessing arXiv research papers. Enables AI agents to search papers with filters for date ranges and categories, download full paper content, read papers in markdown format, and perform semantic search across locally stored papers. Supports citation graph exploration via Semantic Scholar and research alert watches for tracking new publications on topics of interest.

Connect

Add this configuration to .claude/mcp.json

Transport:
{
  "mcpServers": {
    "arxiv": {
      "command": "uvx",
      "args": [
        "arxiv-mcp-server"
      ]
    }
  }
}

Tools (4)

search_papers

Search arXiv papers with optional category, date range, and boolean filters.

{
  "type": "object",
  "required": [
    "query"
  ],
  "properties": {
    "query": {
      "type": "string",
      "description": "Search query (supports boolean operators AND, OR, quotes for exact match)"
    },
    "date_to": {
      "type": "string",
      "description": "Filter papers published before this date (YYYY-MM-DD)"
    },
    "sort_by": {
      "enum": [
        "relevance",
        "date"
      ],
      "type": "string",
      "description": "Sort results by relevance or date"
    },
    "date_from": {
      "type": "string",
      "description": "Filter papers published after this date (YYYY-MM-DD)"
    },
    "categories": {
      "type": "array",
      "items": {
        "type": "string"
      },
      "description": "arXiv category filters (e.g., cs.AI, cs.LG, stat.ML)"
    },
    "max_results": {
      "type": "integer",
      "description": "Maximum number of results to return (default 10, max 50)"
    }
  }
}

download_paper

Download a paper by arXiv ID. Tries HTML first, falls back to PDF.

{
  "type": "object",
  "required": [
    "paper_id"
  ],
  "properties": {
    "start": {
      "type": "integer",
      "description": "Character offset to start reading from (for pagination)"
    },
    "paper_id": {
      "type": "string",
      "description": "arXiv paper ID (e.g., 2401.12345)"
    },
    "max_chars": {
      "type": "integer",
      "description": "Maximum characters to return (for pagination)"
    }
  }
}

read_paper

Read the full text of a locally downloaded paper in markdown format.

{
  "type": "object",
  "required": [
    "paper_id"
  ],
  "properties": {
    "start": {
      "type": "integer",
      "description": "Character offset to start reading from"
    },
    "paper_id": {
      "type": "string",
      "description": "arXiv paper ID of a previously downloaded paper"
    },
    "max_chars": {
      "type": "integer",
      "description": "Maximum characters to return"
    }
  }
}

list_papers

List all papers that have been downloaded locally.

{
  "type": "object",
  "properties": {}
}

Resources

This server does not expose any resources.

Prompts

This server does not expose any prompts.

Server Information

Author
Joe Blazick (@blazickjp)
Repository
https://github.com/blazickjp/arxiv-mcp-server
License
Apache-2.0
Version
0.5.0
Stars
2,800
Last Updated
May 20, 2026
ArXiv MCP Server - MCP Server | OpenModels