Clinical Case Room — Multiplayer AI for clinical teams

The problem

Complex cases are decided by teams, but every tool serves one clinician.

Pathology, imaging, labs, genomics and history arrive in different systems. Before every tumor board or case conference, someone rebuilds the patient's story by hand. Decisions are often recorded late, or never.

Fragmented recordsEach specialist sees part of the picture and may not know what a colleague has ordered.
Manual case preparationCoordinators and physicians spend hours assembling summaries before every meeting.
Single-user AIAI assistants answer one clinician at a time, so the team never shares the same context.
Weak decision recordsRecommendations and who approved them are hard to trace afterwards.

Product

One room per patient. The whole team, the record and the AI in it together.

Each complex patient gets a case room that holds the header, timeline, documents, discussion, decisions, tasks and audit log in one place.

Shared patient memory

Problems, diagnoses, labs, imaging, pathology, treatments and open questions update automatically as documents arrive.

Real-time collaboration

Live discussion with presence, typing indicators and @mentions, synced across every member's browser.

AI for the whole team

Anyone can ask @AI. Answers are visible to everyone and cite the source record behind every statement.

Tumor board briefs and handoffs

The AI drafts briefs, flags missing information and suggests follow-up tasks. Drafts stay drafts until a clinician approves.

Team decision approvals

A decision is final only when every requested reviewer has approved it. The AI can never approve.

Append-only audit log

Every upload, AI action, decision, approval and task change is recorded. The database rejects edits to audit rows.

How it works

From scattered records to a signed-off decision.

Documents arrive

Reports, labs and notes are added to the patient's case room.

Memory updates

The AI extracts problems, findings and open questions into shared patient memory.

The team discusses

Specialists ask @AI and each other. Every AI answer cites its source.

Clinicians decide

Briefs and decisions are approved by the requested reviewers and logged.

Safety by design

Built so the AI supports clinicians and never replaces them.

Evidence firstAI statements cite source records. Fabricated citations are dropped and quotes are checked word for word.
No autonomous clinical actionsThe AI does not diagnose, prescribe, select treatments or triage. Suggested tasks never include medication orders.
Clear content statusEvery item is labelled AI generated, Draft or Human approved.
Role-based approvalOnly authenticated clinicians with the right role can approve, per organization.
Prompt-injection guardDocument and message content is treated as data, never as instructions to the model.
Open and auditableThe full codebase is open source under AGPL-3.0, so hospitals can inspect exactly what runs.

Technology

A real-time collaboration platform with a model-agnostic AI layer.

The AI layer is provider-agnostic, so each hospital can choose where its models run, from a cloud API to its own hardware.

Collaboration platform

Multi-organization workspace with live presence and updates streamed to every member.

Next.jsReactTypeScriptServer-Sent Events

Clinical data layer

Structured patient memory, documents and an append-only audit trail enforced in the database.

PostgreSQLPrismaLISTEN/NOTIFYS3-compatible storage

Evidence-grounded AI

Record organization, cited Q&A, brief drafting and citation verification, behind one provider interface.

LLM provider abstractionCitation verificationOffline demo engine

Next: private, on-premise inference

Many hospitals cannot send patient records to an external API. We are building a deployment option that runs open clinical language models on GPU servers inside the hospital network. This covers record structuring, retrieval over the patient's documents, and brief drafting, so protected health information never leaves the building.

Roadmap

Where Clinical Case Room is today, and where it is going.

Now

Research preview

Open-source MVP with synthetic patients
Multiplayer case rooms, cited AI, approvals, audit log
Offline demo engine that needs no API keys

HIPAA-ready hosted deployment
Pilots with private hospitals and clinics
Evaluation of brief quality and citation accuracy

Later

Hospital-grade deployment

On-premise GPU inference with open models
EHR integration through FHIR and HL7
Specialties beyond oncology: cardiac, spine, transplant

Research

AI for earlier, better-informed cancer decisions.

Clinical Case Room is developed by C&T Technology Global, an AI-for-healthcare company. Alongside the product, we run applied research on early cancer detection.

Physics-informed neural networks for early detection

Ongoing research on mathematical and PINN-based models for early detection of pancreatic cancer from urinary biomarkers.

Grounded AI for multidisciplinary care

Methods to keep language-model output tied to the source record: citation verification, structured patient memory and human-approval workflows.

Every complex patient gets one shared room.