Most bibliographic servers hand back one best match and leave you to hope it is the right person. This one returns several candidates side by side with the institution and citation numbers that actually separate them, which is the step that goes wrong when a common name has a career spanning two labs. Affiliations come back with the years attached, so a move between institutions reads off the record instead of being inferred.
An MCP server over the OpenAlex.org API aimed at one hard problem in academic research: deciding which person an author name refers to. It returns ranked candidates with the institutional and citation context needed to pick between them, then retrieves that author's works with filters for year, venue type, open access and retraction status.
- Multiple ranked candidates for an ambiguous name, each with an institution hint, works count, citation count and an ORCID where one exists, so a match can be chosen on context rather than on spelling — `autocomplete_authors`
- Author records filtered by institution, research topic or country code, carrying name alternatives, affiliations dated by year, h-index and i10-index — `search_authors`
- An author's publications with title, DOI, venue, year, citation count and open-access status, filterable by publication year, work type, institution and retraction, ordered by date or by citations — `retrieve_author_works`
No account and no key: OpenAlex asks only for a contact address, which goes in `OPENALEX_MAILTO`. Python 3.10 or higher. The package is `alex-mcp` (4.8.2 in pyproject), installed from a clone with `pip install -e .` or launched straight from the git URL with uvx. `OPENALEX_MAX_AUTHORS`, `OPENALEX_RATE_PER_SEC` and `OPENALEX_RATE_PER_DAY` are optional and cap how hard it queries the API.
One command — uvx --from git+https://github.com/drAbreu/alex-mcp.git@4.1.0 alex-mcp