Ambient Patient Companion — Data skill for Claude Code

Ambient Patient Companion

**The interface is not designed. It is derived.** `S = f(R, C, P, T)`

A production multi-agent AI health system that continuously generates the optimal clinical interface as a mathematical function of four dynamic variables — Role, Context, Patient State, and Time. Built on Next.js 16, FastMCP Python servers, PostgreSQL, and the Anthropic Claude API.

---

Get Started in 3 Steps

Connect Claude (or any MCP-compatible LLM) to a live clinical intelligence layer and explore insights from your medical reports — imported via HealthEx, your own FHIR data, or the built-in demo patient.

  ┌──────────────────┐     ┌──────────────────┐     ┌──────────────────┐
  │   Step 1         │     │   Step 2         │     │   Step 3         │
  │                  │     │                  │     │                  │
  │  Import this     │────▶│  Get your MCP    │────▶│  Add to Claude   │
  │  repo to Replit  │     │  server URLs     │     │  and explore     │
  │                  │     │                  │     │  your data       │
  └──────────────────┘     └──────────────────┘     └──────────────────┘

---

Step 1 — Import this repository to Replit

[](https://replit.com/new/github/aliomraniH/ambient-patient-companion)

Click the badge above, or do it manually:

1. Go to replit.com → click Create Repl
2. Choose Import from GitHub
3. Paste https://github.com/aliomraniH/ambient-patient-companion
4. Click Import from GitHub — the environment configures automatically

Once imported, open the **Secrets** tab (lock icon in the sidebar) and add:

Secret name	What to put there
ANTHROPIC_API_KEY	Your Claude API key (sk-ant-...)
OPENAI_API_KEY	Your OpenAI key — used by the deliberation engine's GPT-4o critic

Then start all services with one command in the Shell tab:

./start.sh