What it can do
search_papersSearch 8.5M papers by topic, author, or linked entity (gene/protein/disease).
get_paperFull metadata for any arXiv or OpenAlex ID. TLDR, findings, code, models, reviews, links.
similar_papersSPECTER2 nearest neighbors. Papers like this one.
citationsWho cites a paper. Follow-up work across the citation graph.
referencesWhat a paper cites. Walk back through its bibliography.
get_full_textFull body text as Markdown: methods, results, protocols. arXiv, PMC & bioRxiv/medRxiv.
get_figuresReal figure images, labels, and captions extracted from a paper.
get_entitiesLinked genes, proteins, chemicals, diseases, cell lines & clinical-trial IDs.
get_structureResolve a gene/protein to its UniProt accession + 3D AlphaFold structure.
One server. Every client.
Add one URL and Tomesphere shows up as tools inside whatever you already use. Anonymous out of the box — no key to paste.
Not just titles and abstracts.
arXiv + PMC, bioRxiv, medRxiv — every discipline. Title, authors, abstract, venue, citations, OpenAlex ID.
SPECTER2 neighbours + who-cites-what, via OpenAlex.
TLDR, findings, contribution — every paper.
477K UniProt accessions, resolved to their source papers.
146M mentions — genes, proteins, chemicals, diseases.
1.58M papers.
109K → AlphaFold.
402K reviews.
132K NCT-linked.
207K GitHub + HuggingFace links, plus talk videos.
YouTube, SlidesLive.
Free to use.
No payment, no key, nothing to buy. There’s just a light daily rate limit so automated abuse can’t knock the project over.
Add the server and start. Rate-limited by IP.
A higher daily cap when your client connects over OAuth. Still free, still no payment.
One URL. Streamable HTTP.
https://mcp.tomesphere.com/api/mcp