Diagora AI

Not one answer an AI medical board that thinks together.

Comment

Inspiration

Medical decisions are rarely made by one doctor — they are debated, uncertain, and often influenced by cognitive bias. In real clinical settings, multiple specialists collaborate, challenge assumptions, and refine diagnoses before reaching a conclusion.

Most existing AI healthcare tools, however, operate as single-model systems. They provide answers without debate, with limited transparency, and with little visibility into uncertainty.

We asked a simple question:

What if AI could simulate a real clinical board?

This question led to the creation of Diagora.

What it does

Diagora is a multi-agent clinical intelligence system where specialized AI agents — Cardiology, Pulmonology, Neurology, Risk Analysis, and Lab Interpretation — collaborate and challenge each other in real time.

Unlike traditional AI tools that provide a single answer, Diagora makes disagreement visible. It treats uncertainty as a safety signal, not a weakness.

The system:

  • Analyzes symptoms, lab results, and imaging data
  • Simulates structured clinical debate between specialists
  • Identifies hidden risks and conflicting interpretations
  • Surfaces diagnostic uncertainty
  • Generates an explainable, risk-aware clinical report

Diagora transforms AI from a black-box answer into a transparent, collaborative decision-support system.

How we built it

Diagora was built as a modular multi-agent healthcare AI system using Prompt Opinion and a custom product demo interface.

Core architecture:

  • A2A Agent-to-Agent Communication: enables specialist agents to exchange reasoning and challenge each other
  • Orchestrator Agent: coordinates the clinical board, compares opinions, and generates the final report
  • Specialist Agents: Cardiology, Pulmonology, Neurology, Lab Analyst, and Risk Analyst
  • FHIR-inspired data layer: structures patient context for healthcare-style workflows
  • MCP-style tooling concept: supports imaging, lab interpretation, risk assessment, and follow-up workflows

Frontend demo:

  • Next.js App Router
  • TypeScript
  • Tailwind CSS
  • Animated clinical debate interface
  • Doctor and patient report views

Challenges

The biggest challenge was making the system feel like a clinical board rather than a chatbot.

We had to design agents that did not simply agree with each other, but instead challenged assumptions, exposed uncertainty, and reasoned from different clinical perspectives.

Another challenge was presentation. Medical AI can easily become overwhelming, so we designed the interface to make complex reasoning understandable through debate, risk indicators, confidence scores, and structured reports.

Accomplishments that we're proud of

  • Built a working A2A-style clinical board inside Prompt Opinion
  • Created specialized agents for cardiology, pulmonology, neurology, lab analysis, and risk analysis
  • Designed an Orchestrator Agent that detects disagreement and synthesizes final recommendations
  • Built a premium frontend demo showing live agent debate, FHIR-style routing, risk scoring, reports, and follow-up monitoring
  • Made uncertainty visible instead of hiding it behind a single AI answer

What we learned

We learned that the future of AI in healthcare is not about replacing clinicians, but augmenting clinical reasoning.

The most dangerous AI behavior is not always being wrong — it is being confidently incomplete.

Diagora helped us explore how multi-agent systems can improve transparency by showing how different specialists reason, disagree, and converge.

What's next

Next, Diagora could evolve into:

  • A clinical decision-support layer for hospitals and telemedicine platforms
  • A training tool for medical education and diagnostic reasoning
  • A risk detection and second-opinion system for complex cases
  • A follow-up monitoring assistant for post-decision care

Future directions include real FHIR integration, expanded specialist agents, validated medical knowledge bases, and stronger clinician-in-the-loop workflows.

Disclaimer

Diagora is a clinical decision-support prototype. It does not replace licensed medical professionals, provide final diagnoses, or prescribe treatment.

Our goal is to move from AI that simply answers to AI that reasons transparently.

Built With

  • a2a-communication
  • clinical-decision-support
  • fhir
  • google-gemini
  • multi-agent-ai-system
  • next.js
  • prompt-opinion
  • rag
  • tailwind-css
  • typescript
Share this project:

Updates