MediAssist-MCP

MediAssist-MCP is a simple, reliable, hackathon-ready healthcare MCP server built with Python and FastMCP. It exposes synthetic healthcare tools that can plug into MCP clients, the Prompt Opinion Marketplace, and a small Streamlit frontend.

What MCP Is

MCP stands for Model Context Protocol. It is a standard way for AI apps to discover tools, call them with structured arguments, and receive structured JSON back. That makes it a strong fit for hackathon demos where you want interoperability without building a full custom API layer.

Why This Project Works For The Hackathon

AI factor: it exposes multiple healthcare tools an AI agent can call
Potential impact: it covers common demo workflows like triage, medicine lookup, and scheduling
Feasibility: it is local-only, synthetic-only, and easy to run on a MacBook Apple Silicon

What Is Included

Python + FastMCP server
Streamlit frontend for visual testing
Local JSON data source
Modular clean structure
Structured JSON responses
Comments and docstrings
Error handling and logging
Optional local Ollama helper
FHIR-ready notes
SHARP context placeholders
Marketplace publishing notes

Folder Layout

MediAssist-MCP/
├── server.py
├── core/
├── data/
├── demo_script.md
├── docs/
├── fhir/
├── frontend/
├── integrations/
├── services/
├── tools/
├── requirements.txt
├── README.md
└── .env.example

Tools Exposed By MCP

symptom_checker(symptom: str)
emergency_triage(symptoms: str)
bmi_calculator(weight, height)
medicine_info(medicine_name: str)
nutrition_recommendation(condition: str)
appointment_scheduler(name: str, date: str)
mental_health_support(mood: str)
health_risk_assessment(age: int, smoking: bool, diabetes: bool)

All of them use only synthetic healthcare data and simple local logic.

Install

cd "/Users/siva/Documents/Copy 2/New/MediAssist-MCP"
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env

Run Locally

MCP server with stdio transport

MCP_TRANSPORT=stdio python server.py

MCP server with Streamable HTTP transport

MCP_TRANSPORT=streamable-http MCP_HOST=127.0.0.1 MCP_PORT=8000 python server.py

Streamlit frontend

streamlit run frontend/app.py

Test With MCP Inspector

The easiest inspector flow is Streamable HTTP:

Start the MCP server with:

python server.py

Start the Inspector:

npx -y @modelcontextprotocol/inspector

Connect it to:

http://127.0.0.1:8000/mcp

If you prefer stdio, launch the server command directly from the Inspector.

How To Demo For Judges

The best short demo is:

Open the Streamlit app.
Run Symptom checker with fever.
Show Emergency triage with chest pain and shortness of breath.
Show Medicine info with ibuprofen.
Show BMI calculator using the default numbers.
Show Nutrition recommendation for diabetes.
Show Appointment scheduler with a sample name and date.
Finish with Mental health support.

That sequence shows breadth, local reliability, and practical healthcare usefulness.

Prompt Opinion Integration Notes

MediAssist-MCP is designed to be easy to catalog in a Prompt Opinion Marketplace style flow because:

the tool names are stable and readable
every response is structured JSON
no paid API keys are required
no real patient data is used
the server can run over stdio or Streamable HTTP

Suggested integration approach:

Register the MCP server endpoint in Prompt Opinion.
Map each tool name to its schema and sample output.
Use the Streamable HTTP endpoint at http://127.0.0.1:8000/mcp for live demos.
Keep the JSON response shapes stable so agents and marketplace cards stay predictable.

Marketplace Publishing Steps

Confirm the server runs locally with python server.py.
Confirm the frontend works with streamlit run frontend/app.py.
Confirm the tools are visible in MCP Inspector.
Write a short catalog description that explains the synthetic-only healthcare use case.
Publish the MCP endpoint to your marketplace or registry.
Add sample tool calls and response examples for judges and reviewers.

Optional Ollama Support

The project includes integrations/ollama_optional.py as a free/local helper. It is not required for the core demo and is safe to ignore. If you want to experiment with it, set OLLAMA_ENABLED=true in .env.

Local FastMCP Compatibility Layer

This repository includes a bundled mcp/ package that provides a lightweight FastMCP-compatible server implementation. That keeps the project runnable on this MacBook's Python 3.9 interpreter without depending on the external mcp PyPI package, which requires Python 3.10+.

Safety Note

This project is synthetic by design. It is useful for prototyping and hackathon demos, but it is not medical advice and it is not a clinical system.

Terminal Commands

Full local setup

cd "/Users/siva/Documents/Copy 2/New/MediAssist-MCP"
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env

Start the MCP server

python server.py

Start the Streamlit frontend

streamlit run frontend/app.py

Start the MCP Inspector workflow

MCP_TRANSPORT=streamable-http MCP_HOST=127.0.0.1 MCP_PORT=8000 python server.py
npx -y @modelcontextprotocol/inspector

Demo Assets

See demo_script.md for a short live demo script, judge-ready talking points, and copy-paste commands.

MCP Servers