arxiv-dl

A command-line tool for downloading, naming, and cataloging research papers from arXiv and major paper repositories.

arxiv-dl is an independent project and is not affiliated with arXiv.

Features

• Accepts arXiv IDs, individual paper URLs, and supported listing or collection pages.See Supported Inputs for the complete list of supported targets.
• Saves consistently named PDFs, optional text or Markdown notes, and a local JSON paper index.
• Extracts available metadata such as title, authors, abstract, venue, year, comments, and BibTeX. Availability varies by source.
• Supports a configurable download directory and optional parallel downloads with aria2.

Installation

arxiv-dl requires Python 3.9 or newer. Install it with pipx to keep it isolated from other Python packages:

If pipx is already installed:

pipx install arxiv-dl
paper --help

Current pipx releases require Python 3.10 or newer. If pipx is not installed, use the setup command for your platform, restart the terminal, and then run the commands above.

Platform	Install pipx
macOS	brew install pipx pipx ensurepath
Debian 12+ / Ubuntu 23.04+	sudo apt update sudo apt install pipx pipx ensurepath
Fedora	sudo dnf install pipx pipx ensurepath
Windows PowerShell	py -m pip install --user pipx py -m pipx ensurepath

On Windows, replace py with python or python3 if needed. See the pipx installation guide for other platforms.

Upgrade an existing installation with:

pipx upgrade arxiv-dl

Install with pip in a virtual environment

Use this method if you have Python 3.9 or prefer not to use pipx.

On macOS or Linux:

python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade arxiv-dl
paper --help

On Windows PowerShell:

py -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install --upgrade arxiv-dl
paper --help

On Command Prompt, use .venv\Scripts\activate.bat instead of the PowerShell activation command. The paper command is available while the virtual environment is active.

If PowerShell blocks script activation, use the environment directly:

.\.venv\Scripts\python.exe -m pip install --upgrade arxiv-dl
.\.venv\Scripts\paper.exe --help

Usage

paper [OPTIONS] TARGET [TARGET ...]

arxiv-dl and getpaper are alternative command names for paper.

Examples

# Download one paper by arXiv ID
paper 1512.03385

# Download multiple targets
paper 2103.15538 https://arxiv.org/abs/1512.03385

# Download all papers from the current Hugging Face daily listing
paper https://huggingface.co/papers

# Choose an output directory and skip the notes file
paper 1512.03385 --download-dir ./papers --pdf-only

Supported inputs

Source	Accepted input
arXiv	Modern or legacy arXiv IDs; abstract, PDF, and HTML URLs
alphaXiv	Paper routes containing a valid arXiv ID; resolved through canonical arXiv URLs
Hugging Face Papers	Individual papers; daily, weekly, monthly, trending, user, and organization listings; collections
ICLR	Proceedings abstract and PDF URLs, including short-route aliases
NeurIPS / NIPS	Abstract and PDF URLs from the current proceedings or legacy NIPS site
CVF Open Access	Abstract and PDF URLs for CVPR, ICCV, WACV, and ACCV, including workshops and findings tracks
ECVA	ECCV abstract pages; direct PDF support varies by year, so prefer the abstract page

View detailed input formats

✅ Supported, 🚧 not yet supported, ⚠️ limited support

• arXiv
  • ✅ arXiv ID: 1512.03385 or arXiv:1512.03385
  • ✅ Legacy arXiv ID: alg-geom/9708001 or cs/0002001
  • ✅ Abstract URL: https://arxiv.org/abs/1512.03385
  • ✅ PDF URL: https://arxiv.org/pdf/1512.03385.pdf
  • ✅ HTML URL: https://arxiv.org/html/2506.15442
• alphaXiv
  • ✅ Abstract URL: https://www.alphaxiv.org/abs/2312.16682v2
  • ✅ PDF URL: https://www.alphaxiv.org/pdf/2312.16682v2
  • ✅ Other paper routes containing an arXiv ID, including overview, HTML, localized, Markdown, and direct PDF asset URLs
  • alphaXiv targets resolve through the corresponding canonical arXiv URLs.
• Hugging Face Papers
  • ✅ Individual paper: https://huggingface.co/papers/2605.12357
  • ✅ Current daily papers: https://huggingface.co/papers
  • ✅ Daily papers: https://huggingface.co/papers/date/2026-05-22
  • ✅ Weekly papers: https://huggingface.co/papers/week/2026-W21
  • ✅ Monthly papers: https://huggingface.co/papers/month/2026-05
  • ✅ Trending papers: https://huggingface.co/papers/trending
  • ✅ User or organization papers: https://huggingface.co/huggingface/papers
  • ✅ Collection: https://huggingface.co/collections/Testerpce/memory
• CVF Open Access (CVPR, ICCV, WACV, ACCV)
  • ✅ Abstract URL: https://openaccess.thecvf.com/content/**/html/**/*.html
  • ✅ PDF URL: https://openaccess.thecvf.com/content/**/papers/**/*.pdf
  • ✅ Workshop and findings-track URLs
• ECVA (ECCV)
  • ✅ Abstract URL: https://www.ecva.net/papers/eccv_<year>/papers_ECCV/html/<paper>.php
  • ⚠️ Direct PDF support varies by year; prefer the abstract URL.
• NeurIPS Proceedings / NIPS legacy site
  • ✅ NeurIPS abstract URL: https://proceedings.neurips.cc/paper_files/paper/**/hash/**/*.html
  • ✅ NeurIPS PDF URL: https://proceedings.neurips.cc/paper_files/paper/**/file/**/*.pdf
  • ✅ NIPS abstract URL: https://papers.nips.cc/paper_files/paper/**/hash/**/*.html
  • ✅ NIPS PDF URL: https://papers.nips.cc/paper_files/paper/**/file/**/*.pdf
• ICLR Proceedings
  • ✅ Abstract URL: https://proceedings.iclr.cc/paper_files/paper/**/hash/**/*-Abstract-Conference.html
  • ✅ PDF URL: https://proceedings.iclr.cc/paper_files/paper/**/file/**/*-Paper-Conference.pdf
  • ✅ Short /paper/** routes and -Abstract.html aliases, normalized to the canonical HTTPS -Conference URLs
• OpenReview
  • 🚧 Not yet supported

Options

Option	Description
-d, --download-dir DIR	Set the directory for this run; overrides the environment variable and default.
-p, --pdf-only	Download the PDF without creating a notes file.
--notes-format {txt,md}	Set the notes format (default: txt).
-n, --n-threads N	Request 1–16 download connections (default: 1). Values above 1 use aria2 when available; CVF uses one connection.
-v, --verbose	Show full details.
--verbose-level LEVEL	Set output to silent, minimal, default, or verbose.
--skip-update-check	Skip the package update check.

Run paper --help for the full command reference.

Faster downloads with aria2

aria2 is optional. Install it, ensure aria2c is on your PATH, and request more than one connection:

Platform	Install aria2
macOS	brew install aria2
Debian / Ubuntu	sudo apt install aria2
Fedora	sudo dnf install aria2
Windows PowerShell	winget install --exact --id aria2.aria2

paper --n-threads 5 1512.03385

Configuration

Papers are saved to ~/Downloads/ArXiv_Papers by default (%USERPROFILE%\Downloads\ArXiv_Papers on Windows). Use --download-dir for a single run or set ARXIV_DOWNLOAD_FOLDER for a persistent default.

On macOS or Linux, add this to .bashrc, .zshrc, or the relevant shell profile:

export ARXIV_DOWNLOAD_FOLDER="$HOME/Documents/Papers"

Reload the shell profile or open a new terminal for the change to take effect.

On Windows PowerShell, set a persistent user environment variable and open a new terminal:

[Environment]::SetEnvironmentVariable("ARXIV_DOWNLOAD_FOLDER", "$HOME\Documents\Papers", "User")

The resolution order is --download-dir, ARXIV_DOWNLOAD_FOLDER, then the default directory.

Python API

from arxiv_dl import download_paper

download_paper(
    target="1512.03385",
    download_dir=".",
    set_verbose_level="silent",
)

Contributing

See DEVELOPMENT.md for development, testing, build, and release instructions.

License

MIT © 2021–2026 Mark H. Huang.