litsurvey: scientific literature search and prior-art checks

litsurvey lets your agent search the scientific literature properly instead of guessing from the web. The skill drives the free litsurvey command-line tool.

What it does

• Searches OpenAlex, Semantic Scholar, arXiv, PubMed, TechRxiv, Research Square and the IACR ePrint archive in one go, removes duplicates, and ranks the merged list. Europe PMC and Crossref can be added.
• For any paper, given by title, DOI or arXiv id: who cites it (most-cited first), what it cites, similar papers, its full record and BibTeX, and a legal open-access copy.
• research "question": a state-of-the-art survey organised by theme, with numbered citations taken from live search results.
• novelty "claim": the closest prior work and a verdict (clearly novel / incremental / substantially anticipated / cannot determine).
• Every report ends with a log of the queries that were run and a check of each cited DOI and arXiv id against the search results.
• Exports to BibTeX, RIS, CSV, JSON or Markdown.

Before you install this skill

The skill needs the litsurvey tool, which you install once (Python 3.10 or newer, no other dependencies):

pipx install litsurvey

or pip install litsurvey. Check it with litsurvey --version. A free Semantic Scholar API key is recommended; litsurvey init stores it.

Which LLM it uses

Searching and citation lookups need no LLM at all. For research and novelty you choose where the model runs: your Claude, ChatGPT or Gemini subscription through its command-line agent, a local model through Ollama (nothing leaves your machine), or an API key.

Privacy

Search queries go to the public scholarly APIs. With a local model, your claim or question stays on your machine. Details: https://github.com/udaykdk/litsurvey/blob/main/docs/confidentiality.md

Links

• Source and documentation: https://github.com/udaykdk/litsurvey
• PyPI: https://pypi.org/project/litsurvey/
• Easiest way to try it: run litsurvey web for a browser page with every feature.

Licence and author

MIT licence. Designed by Uday Khankhoje, Department of Electrical Engineering, IIT Madras. Bug reports: https://github.com/udaykdk/litsurvey/issues