This release makes OpenOnco usable and discoverable by LLMs, so an assistant can route an oncology question through the deterministic, source-cited engine instead of answering from memory.
Highlights
- MCP server (
mcp_server/) — expose the rule engine to any Model Context Protocol client (Claude Desktop, Cursor, …). Tools:engine_info,list_diseases,generate_treatment_plan,generate_diagnostic_brief. The LLM relays cited engine output; it never picks the regimen or dose (CHARTER §8.3). Runs locally — patient data stays on the machine. - LLM discoverability —
llms.txt+ newllms-full.txt, and richer JSON-LD (license, keywords, GitHubsameAs,SoftwareApplication) on every page of openonco.info. - Root
LICENSE(MIT) added so the license is detectable; content/specs remain CC BY 4.0.
Knowledge base (state 2026-06-17)
92 diseases · 664 indications · 384 regimens · 298 drugs · 594 red flags · 444 cited sources · 16 virtual MDT skills.
Honest about maturity
Early-stage v0.1. Most clinical content is STUB — only 15 of 806 entities have two-Clinical-Co-Lead sign-off; the rest are "proposed, not approved." There is no formal clinical validation study. Treat every output as a draft to verify.
Links
- Site / demo: https://openonco.info · https://openonco.info/try.html
- MCP server: https://github.com/romeo111/OpenOnco/tree/main/mcp_server
OpenOnco is an informational clinical decision support resource for healthcare professionals — not a medical device, not FDA-cleared, not clinically validated, and not for direct patient use. Every recommendation must be verified by a qualified oncologist.