The agent for research teams

Plans the approach, queries domain sources in parallel, returns a cited brief — dozens of tool calls, zero babysitting.

arXiv, PubMed, Semantic Scholar, bioRxiv queried at once — ranked evidence with inline citations, not a list of links.

Writes and runs Python in a serverless sandbox — figures and tables come back inline, every input cited.

AlphaFold / PDB structures, ClinVar variants, Open Targets — fetched and visualized inside the conversation.

Recruiting trials, patent landscapes, regulatory guidance — matched records with IDs and the primary source.

Drop in reports — extracts what they state, preserves negations, links findings to cited literature and coding systems.

Molecular structures, lightweight charts, LaTeX, and elevated citation cards rendered next to the prose.

The right safety frame per question — clinical caveats, label limits, jurisdiction warnings, or none at all.

OpenAI-compatible endpoint to drop Cortexa into EHRs, notebooks, and pipelines. Metered, zero lock-in.