fastpaper（fastpaper 搜索）

fastpaper

Fast academic paper search, download & read from 17 sources.

fastpaper is a pre-installed standalone CLI. Before first use, verify it is available: which fastpaper. It is NOT a Python package — do NOT attempt to install via pip.

Source selection by domain

Pick sources based on the user's research domain:

CS / AI / Math / Physics

• arxiv — Cornell preprint archive; physics, math, CS, statistics, EE, quantitative biology/finance, economics. search, download, read
• dblp — CS-focused bibliography index. search

Biomedical / Life sciences

• pubmed — NLM index, 35M+ citations, biomedical and life sciences abstracts. search
• pmc — NLM full-text archive of peer-reviewed biomedical and life sciences literature. search, download, read
• europepmc — Life sciences superset of PMC by EMBL-EBI; adds patents, preprints, clinical guidelines. search, download, read
• biorxiv — Life sciences preprints by CSHL. search, download, read
• medrxiv — Medical/health science preprints by CSHL. search, download, read

Cross-discipline / Broad coverage

• semantic — Allen AI, AI-powered semantic search + citation graph, all disciplines. search, download, read
• crossref — DOI registry, metadata queries across all disciplines. search
• openalex — Open index (successor to MS Academic Graph), 200M+ works. search
• scholar — Google Scholar, broadest coverage (experimental, rate-limited). search

Open access aggregators

• core — Largest global OA aggregator, full text from institutional repos and journals. search, download, read
• openaire — EU open science infrastructure, aggregates worldwide OA research. search, download, read
• doaj — Directory of quality-reviewed OA journals, all subjects. search, download, read
• unpaywall — OA link resolver by DOI, finds legal free versions (needs UNPAYWALL_EMAIL). search

Open repositories

• zenodo — CERN/OpenAIRE general-purpose repository (datasets, software, papers), all disciplines. search, download, read
• hal — French national multi-disciplinary open archive by CNRS (some embargo periods). search, download, read

When you have a paper ID or DOI

Auto-detect source and fetch directly:

fastpaper get <DOI|arXiv_ID|PMID|URL>

When you need to search

fastpaper search <source> <query>

For broad topics, search multiple sources in parallel:

fastpaper search arxiv "transformer attention" --format json &
fastpaper search semantic "transformer attention" --format json &
wait

Each process is independent; failures don't affect other sources.

When you need full text or specific sections

fastpaper read <source> <id>                          # full text
fastpaper read <source> <id> --metadata-only          # metadata only
fastpaper read <source> <id> --section <SEC>          # extract section
fastpaper read local ./file.pdf                       # local PDF

Sections: abstract, introduction, methods, results, discussion, conclusion, references, full (default).

When you need to download PDF

fastpaper download <source> <id>

Download-capable sources: arxiv, pmc, semantic, biorxiv, medrxiv, europepmc, core, doaj, zenodo, hal, openaire.

All output uses --format json for structured agent consumption.

Exit code 0 = success, non-zero = error (see --help for codes).