arxiv-cli

A Rust CLI for searching arXiv, listing the latest papers in a category, downloading papers, and managing a local paper library.

Install

This project is not published to crates.io yet. Install it from source:

git clone <your-repo-url> arxiv-cli
cd arxiv-cli
cargo install --path .

or install it from npm:

npm install -g arxiv-cli

Verify the install:

arxiv --help

Run without global install during development:

cargo run -- --help

What It Can Do

• Search arXiv by keyword, title, author, category, and submitted date range
• List the newest papers in a category, always sorted by submission time
• Show full paper metadata including abstract, PDF URL, and source URL
• Download PDF, LaTeX source, or both
• Automatically extract downloaded source archives by default
• Register a default download directory
• Download multiple papers in parallel
• Maintain a local JSON-backed library
• Output either human-readable tables or JSON

Quick Start

# keyword search
arxiv search "diffusion models"

# title-only search
arxiv search --title "skill"

# search with filters
arxiv search "transformer" --author "Vaswani" --category cs.CL --from 2024-01-01 --to 2024-12-31

# list newest papers in a category
arxiv latest cs.CL --limit 10

# show one paper
arxiv show 1706.03762

# download PDF and source
arxiv download 1706.03762 2401.12345 --format both --jobs 4

# set a default download directory
arxiv config set-download-dir ~/Documents/papers/arxiv

Command Guide

arxiv search

Use search when you want flexible retrieval across general arXiv text fields.

Examples:

arxiv search "skill"
arxiv search "diffusion models" --limit 20
arxiv search "llm" --category cs.CL --sort submitted
arxiv search --title "large language model"
arxiv search --author "Yann LeCun"
arxiv search "transformer" --from 2025-01-01 --to 2025-12-31
arxiv search "skill" --include-abstract
arxiv search "skill" --json
arxiv search "skill" --json --include-abstract

Key parameters:

• QUERY: optional keyword query. Multiple words are combined with AND
• --title <TITLE>: search title only
• --author <AUTHOR>: filter by author name
• --category <CATEGORY>: filter by arXiv category such as cs.CL, cs.LG, math.PR
• --from <YYYY-MM-DD> / --to <YYYY-MM-DD>: submitted-date range
• --sort relevance|updated|submitted: choose ranking field
• --order asc|desc: choose sort direction
• --limit <N>: number of results to return
• --start <N>: pagination offset
• --include-abstract: include abstracts in table output and JSON output
• --json: print machine-readable JSON. By default JSON omits abstract_text; add --include-abstract to include it

Search behavior:

• arxiv search "diffusion models" means diffusion AND models
• arxiv search --title "diffusion models" searches that phrase in the title field
• Plain keyword search uses arXiv's broad text search fields, which include title/abstract-style matching

arxiv latest

Use latest when you already know the category and only want the newest submissions.

Examples:

arxiv latest cs.CL
arxiv latest cs.CL --limit 20
arxiv latest cs.CL --from 2025-01-01 --to 2025-12-31
arxiv latest math.PR --include-abstract
arxiv latest cs.LG --json
arxiv latest cs.LG --json --include-abstract

Key parameters:

• <CATEGORY>: required arXiv category, such as cs.CL, cs.LG, stat.ML
• --limit <N>: number of papers to return, default 10
• --from <YYYY-MM-DD> / --to <YYYY-MM-DD>: submitted-date range
• --include-abstract: include abstracts in table output and JSON output
• --json: print JSON instead of a table. By default JSON omits abstract_text; add --include-abstract to include it

Behavior:

• latest is always sorted by submitted time descending
• It is designed for category feeds, not institution/affiliation search

arxiv show

Use show to inspect one paper in detail.

arxiv show 1706.03762
arxiv show 1706.03762 --json

Returned information includes:

• title
• authors
• abstract
• categories
• published / updated time
• PDF URL
• source URL
• local library status if the paper is already saved

arxiv download

Use download to save PDF, source, or both to disk.

Examples:

arxiv download 1706.03762
arxiv download 1706.03762 --format source
arxiv download 1706.03762 2401.12345 --format both --jobs 4
arxiv download 1706.03762 --output ~/Downloads/arxiv
arxiv download 1706.03762 --force
arxiv download 1706.03762 --no-library-update

Key parameters:

• [IDS]...: one or more arXiv IDs
• --format pdf|source|both: what to download
• --output <DIR>: override the default download directory for this command
• --jobs <N>: parallel download concurrency, default 4
• --force: re-download even if the file already exists
• --no-library-update: skip writing download status back to the local library

Output behavior:

• successful downloads print the paper ID
• the CLI also prints the saved file path for PDF and/or source
• source downloads are extracted by default; the printed source: path is the extracted directory

Example output:

downloaded 1706.03762v7
  pdf: /.../1706.03762v7.pdf
  source: /.../arXiv-1706.03762v7 # tar.gz is extracted to this directory automatically

arxiv library

Use library to maintain a local index of papers you care about.

Examples:

arxiv library add 1706.03762
arxiv library list
arxiv library list --downloaded-only
arxiv library list --category cs.CL
arxiv library list --author "Vaswani"
arxiv library show 1706.03762
arxiv library remove 1706.03762
arxiv library remove 1706.03762 --purge-files

Subcommands:

• library add <ID...>: save metadata into the local library without downloading
• library list: list library entries
• library show <ID>: show one local entry with file paths and status
• library remove <ID>: remove one entry from the index
• --purge-files: remove associated local files too

arxiv config and arxiv path

Use config to inspect or update default behavior.

Examples:

arxiv config show
arxiv config set-download-dir ~/Documents/papers/arxiv
arxiv path

Use cases:

• register a default download folder once, then omit --output
• print resolved config/data/library/download paths for scripting

Common Workflows

1. Search then inspect

arxiv search "test-time scaling" --category cs.CL --limit 5
arxiv show 2501.12345

2. Follow the latest papers in one category

arxiv latest cs.CL --limit 20
arxiv latest cs.CL --from 2026-01-01 --to 2026-04-01

3. Download a paper set for local reading

arxiv config set-download-dir ~/Documents/papers/arxiv
arxiv download 1706.03762 2401.12345 --format both --jobs 4

4. Build a small local paper library

arxiv library add 1706.03762 2401.12345
arxiv library list
arxiv library show 1706.03762

Storage

By default the CLI stores:

• config.toml in the platform config directory
• library.json in the platform data directory
• downloaded files under papers/<arxiv_id>/

Print the resolved paths with:

arxiv path

Notes

• latest currently works on arXiv categories, not institution/affiliation feeds
• Search date filters apply to submitted time
• If you want script-friendly output, prefer --json

Test

cargo test