Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@MedFlow MCP Schedule a follow-up for patient John Doe next Tuesday."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
MedFlow MCP is an AI-powered Hospital Operations server built with NitroStack. It serves as the "brain" for modern healthcare management, allowing AI assistants (like Claude, Cursor, and Nitro Studio) to seamlessly interact with hospital databases, trigger automated alerts, schedule appointments, and manage maintenance via Jira.
Built for the NitroStack MCP Hackathon, this project demonstrates how the Model Context Protocol (MCP) can completely revolutionize complex operational workflows in a healthcare setting.
🎫 Automated Jira Integration: AI can automatically detect malfunctioning hospital equipment (e.g., MRI machines, Defibrillators) from natural language and instantly generate a formatted Jira ticket for the maintenance team via the Atlassian API.
🎨 Glassmorphism UI Widgets: Includes rich, beautifully styled Next.js frontend widgets that pop up natively inside the AI chat interface (Patient Profiles, Doctor Dashboards, and Surgery Readiness checklists).
🩺 Patient & Appointment Management: Exposes AI tools for scheduling follow-ups, prioritizing critical patients, and viewing real-time doctor queues.
🚨 Emergency Alerting System: Tools to broadcast critical CODE emergencies directly to ER teams via Webhook (Slack/Discord).
🧠 Diagnostic Vision AI (Mock): A tool designed to read X-rays and MRI scan notes to provide instant clinical AI impressions.
Related MCP server: Personal JIRA MCP
Framework: NitroStack (Official MCP Framework)
Backend: Node.js, TypeScript
Frontend (Widgets): Next.js, React, Tailwind CSS, Framer Motion
Integrations: Atlassian Jira API v3 (REST)
git clone https://github.com/tsjharsha/MedFlowMCP-AI-.git
cd MedFlowMCP-AI-npm install
npm --prefix src/widgets installCreate a .env file in the root directory and add your Jira API credentials:
JIRA_DOMAIN=your-team.atlassian.net
JIRA_PROJECT_KEY=SCRUM
JIRA_EMAIL=your-email@domain.com
JIRA_API_TOKEN=your-jira-api-tokenStart the development server with hot-reloading for both the backend MCP server and frontend widgets:
npm run devBecause this is an MCP Server, you don't interact with it directly through a browser. You connect an AI to it!
Nitro Studio: Start your local server, open Nitro Studio, and connect to http://localhost:3000/mcp.
Cursor IDE: Go to Settings -> Features -> MCP -> Add New. Select SSE, name it MedFlow, and set the URL to your deployed NitroCloud URL or http://localhost:3000/mcp.
Once connected, simply talk to the AI (e.g., "Show me the dashboard for Doctor Alice" or "Report that the X-Ray in the lab is broken") and watch it use the tools and render the widgets!
Built with ❤️ for the NitroStack Hackathon.
Send a medical scan (X-Ray, MRI) to a specialized Diagnostic Vision AI model for anomaly detection.
Inputs: scanType (required), patientId (required), clinicalNotes (required).
Extract structured medical notes (Symptoms, Diagnosis, Medicines, Tests, Advice) from a raw doctor voice transcript.
Inputs: transcript (required).
Fetch the medical history of a patient by ID.
Inputs: patientId (required).
Get the current queue for a specific doctor.
Inputs: doctorId (required).
Send a request for lab tests to the laboratory.
Inputs: tests (required), doctorId (required), patientId (required), priority.
Reorder a doctor's queue by changing an appointment priority.
Inputs: newPriority (required), appointmentId (required).
Save finalized consultation notes to the patient record.
Inputs: notes (required), doctorId (required), patientId (required), appointmentId (required).
Register a new patient or update an existing one.
Inputs: age (required), name (required), gender (required), contact (required), +1 more.
Schedule a new appointment for a patient with a doctor.
Inputs: time (required), type (required), doctorId (required), complaint (required), +1 more.
Schedule a follow-up appointment in a given number of days.
Inputs: doctorId (required), patientId (required), daysFromNow (required).
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 10 tools
Each tool targets a distinct action or resource: patient registration, history retrieval, appointment scheduling, queue management, follow-up scheduling, queue prioritization, consultation recording, note generation, lab tests, and image analysis. No significant overlap.
All tool names follow a consistent verb_noun pattern in snake_case (e.g., register_patient, schedule_appointment, analyze_medical_image). No mixing of conventions.
10 tools cover the main workflows of a medical practice: patient management, appointments, queue, consultations, lab, and imaging. Not too few or too many.
Covers core clinical operations well, including patient history, appointments, queue, consultation notes, lab tests, and image analysis. Minor gaps like billing or discharge are absent but not critical for the apparent scope.
Hosted MCP server exposing US hospital procedure cost data to AI assistants
AI-powered incident management and server monitoring via MCP.
MCP server unifying ERPs, CRMs, APIs and knowledge base for Claude, ChatGPT and Gemini.
Hosted MCP server for Cliniko — patients, appointments, availability, and invoices for AI agents.
