It covers the narrow thing a model actually needs mid-conversation: find papers on a subject, optionally within a date range, then fetch the full record for one you already have a DOI for. Because the key variables in the config example are not wired up yet, it runs against the public endpoints as-is.
A Python MCP server, built on FastMCP and httpx, that searches academic papers across Semantic Scholar and Crossref and returns metadata, abstracts and full-text links where they exist. The README notes the software is under active development and that features and functionality are subject to change.
- Paper search across the sources by free-text query, with an optional result limit that defaults to 10 — `search_papers`
- Topic search with an optional start and end year, the query capped at 300 characters, returning titles, authors, years, abstracts, TL;DR summaries where available, venue and open access information — `search_by_topic`
- A full record for one paper by DOI or Semantic Scholar ID, choosing `crossref` or `semantic_scholar` as the source (`crossref` by default) — `fetch_paper_details`
- Detail responses carrying title, authors, year, DOI, venue, open access status and PDF URL (Semantic Scholar only), plus abstract and TL;DR when available
Python with `uv`: `uv add "mcp[cli]" httpx`, then run the server with `uv run server.py` and point the client's command at that path. The data sources are the Semantic Scholar API and the Crossref API. The README's env block shows `SEMANTIC_SCHOLAR_API_KEY` and `CROSSREF_API_KEY` and marks them "These are not actually implemented" — so leave them out and use the standalone instructions. There is also a Smithery one-liner, `npx -y @smithery/cli install @afrise/academic-search-mcp-server --client claude`, which the README describes as largely untested and suggests skipping until it is fixed. The package declares itself `academic-search` (0.1.0). AGPL-3.0, which means network use of a modified copy obliges you to publish the source.
One command — uvx academic-search