This server downloads academic papers and retrieves their metadata using DOI, arXiv ID, or URL identifiers.
Download papers (
paper_download): Download 1–50 papers per call with configurable concurrency (default 10 parallel workers), optional PDF-to-Markdown conversion, and customizable output directoriesRetrieve paper metadata (
paper_get_metadata): Fetch metadata (title, authors, year, journal, OA status, available sources) without downloading the PDF, using Unpaywall, Crossref, and arXiv APIsSmart source prioritization: Automatically tries open-access sources (Unpaywall, arXiv, CORE) before falling back to Sci-Hub as a last resort
Flexible input: Accepts DOI, arXiv ID, or direct URLs
Configuration: Requires an email via the
PAPER_DOWNLOAD_EMAILenvironment variable for the Unpaywall API
Enables the automatic identification and downloading of academic preprints directly from the arXiv repository.
Allows for the retrieval of academic papers and associated metadata using Digital Object Identifiers (DOIs) from various scholarly sources.
Supports downloading academic research articles by integrating with PubMed Central (PMC) for open access content.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Paper Download MCP ServerDownload the paper with DOI 10.1038/nature12373"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Paper Download MCP Server
English | 简体中文
MCP server for downloading academic papers by DOI, arXiv ID, or URL.
What You Get
paper_download: Download one or more papers (1-50 per call)paper_get_metadata: Get paper metadata without downloadingOptional PDF-to-Markdown conversion via
to_markdown
Quick Start (MCP Clients)
Before configuration, make sure uvx is available:
Claude Code
Add as a project-scoped MCP server:
This writes .mcp.json in the current project. Equivalent config:
Codex
Add with CLI:
Equivalent ~/.codex/config.toml snippet:
Claude Desktop
Edit Claude Desktop MCP config:
macOS:
~/Library/Application Support/Claude/claude_desktop_config.jsonWindows:
%APPDATA%\\Claude\\claude_desktop_config.json
Restart Claude Desktop after editing the file.
Configuration
Required
PAPER_DOWNLOAD_EMAIL: Required for Unpaywall API usage.
Optional (Advanced)
PAPER_DOWNLOAD_OUTPUT_DIR: Global fallback output directory.
In most cases, you do not need PAPER_DOWNLOAD_OUTPUT_DIR. Prefer passing output_dir in the paper_download tool call when you want a specific location.
Legacy env vars are still supported for compatibility:
SCIHUB_CLI_EMAILSCIHUB_OUTPUT_DIR
Tools
paper_download
Download papers with configurable concurrency (default parallel=10).
If parallel=1, papers are processed sequentially with a 2-second delay between items.
Parameters:
identifiers(required):list[str], 1-50 itemsoutput_dir(optional): target directory (default./downloads)parallel(optional): concurrent workers,1-50(default10)to_markdown(optional): convert PDF to Markdown (falseby default)md_output_dir(optional): Markdown directory (default<output_dir>/md)
Examples:
paper_get_metadata
Get metadata quickly (no PDF download).
Parameters:
identifier(required): DOI, arXiv ID, or URL
Example:
Troubleshooting
PAPER_DOWNLOAD_EMAIL environment variable is required
Set PAPER_DOWNLOAD_EMAIL in your MCP server config.
uvx: command not found
Install uv, then re-run the MCP configuration.
Download path errors
Pass a writable directory with output_dir, for example:
Legal Notice
This tool can access papers from multiple sources, including Unpaywall and Sci-Hub. You are responsible for complying with copyright and local laws in your jurisdiction.
License
MIT. See LICENSE.