What is paper-search-cli?

Paper Search CLI is a Node.js command line tool for finding, validating, and downloading academic papers. It is aimed at researchers, students, and increasingly at automated agents that need to look up scholarly literature. The tool started life as an earlier project called Paper Search MCP, which ran as a long-living background service. This version takes the same broad coverage of academic sources and the same unified data model, but repackages it as a plain CLI: each command starts, returns a result, and exits. There is nothing to keep alive. The headline number is 25 academic sources. The list includes Crossref, OpenAlex, PubMed, PubMed Central, Europe PMC, arXiv, bioRxiv, medRxiv, Semantic Scholar, CORE, OpenAIRE, DBLP, ACM Digital Library metadata, USENIX metadata, OpenReview, Web of Science, Google Scholar, IACR ePrint, Sci-Hub, IEEE Xplore, ScienceDirect, Springer Nature, Wiley, Scopus, and Unpaywall. Some of these only provide metadata, some can serve full text PDFs, and some need an API key. The README is careful to document which capability each source supports through a detailed table. The core design goals are stated up front. Free public metadata and open access full text routes are tried before restricted ones. Output is machine-readable JSON on stdout so an agent or script can parse the results, while human-readable diagnostics go to stderr. A funnel-style fallback download tries the native source first, then discovered PDF URLs, then open access aggregators such as PMC and Unpaywall, and finally Sci-Hub as a last resort unless the user disables that with useSciHub=false. Rate limits and retries are handled per platform. Installation needs Node.js 18 or newer. The typical sequence is to install the package globally with npm, run paper-search setup to write any API keys and contact emails into the user config, and then start searching. The README shows a quick example: paper-search search "machine learning" --platform crossref --max-results 3 --pretty. Other useful commands are status, tools, config doctor, and run, which is the precise advanced call for invoking a specific tool by name. The project is MIT licensed and at version 0.1.2. The author thanks the LinuxDo community for shaping the CLI plus Skill direction and the paper search workflow. Platforms are grouped into broad families such as general scholarly metadata, medicine and life sciences, preprints, computer science, open full text, citation indexes, and DOI-targeted lookup. The full README is longer than what was shown.