This HMCP server implements the Healthcare Model Context Protocol (HMCP), an healthcare-specific extension of MCP designed to securely enable interoperability between healthcare AI agents. It supports agent-to-agent sampling endpoints, OAuth 2.0-based authentication following SMART on FHIR, and transport options including SSE and streamable HTTP for real-time or streaming interactions. Use the provided HMCPServer to expose a sampling endpoint for inter-agent communication and connect with HMCPClientConnector from client code to discover tools, send messages, and coordinate multi-agent workflows. Typical usage involves launching the Python-based server, then connecting a client to exchange messages, query tools, and run multi-agent tasks such as updating EMR data or retrieving patient context while maintaining access controls and audit trails.

Once running, you can connect clients with HMCPClientConnector, list available tools, and send messages to agents via the simplified interface. The server supports both single-agent and multi-agent workflows, allowing complex collaboration across healthcare agents (e.g., EMR data agents, patient data agents). Tools can be discovered and invoked through the client, enabling actions like fetching patient identifiers, updating clinical data, or querying patient context. The sampling endpoint demonstrates how agents process incoming messages and return structured results suitable for downstream processing in healthcare AI pipelines.

Prerequisites

• Python 3.8+ (recommended)
• Pipenv or virtual environment support (optional but recommended)
• Access to install Python packages from PyPI

Installation steps

1. Clone or download the repository: git clone <repo-url> cd innovaccer-healthcare-mcp

2. Create and activate a Python environment (optional but recommended): python -m venv venv

   Windows

   venv\Scripts\activate

   macOS/Linux

   source venv/bin/activate

3. Install dependencies (adjust as per your project setup): pip install -r requirements.txt

4. Build or install HMCP package if required by the project setup (example shown in Quick Start):

   Temporary steps till the package isn't published:

   pip install hatch hatch build

5. Run the HMCP server (example command from the README): python -m hmcp.server.hmcp_server

Note: Depending on your deployment, you may run this via a container or a process manager. Adjust environment variables and port bindings as needed.

Tips and notes:

• Transport options: The HMCP server supports both SSE and streamable-http transports. Choose the one that fits your network security and firewall constraints.
• Authentication: HMCP uses OAuth 2.0 in line with SMART on FHIR. Ensure proper client credentials and token handling are configured in your environment.
• Tool discovery: Use the HMCPClientConnector to automatically discover tools and resources exposed by the server. This simplifies integration with other agents.
• Multi-agent workflows: HMCP is designed for coordinating multiple healthcare agents. You can chain messages across EMR, patient data, and analytics agents to complete complex tasks.
• Auditing and security: Ensure log and audit trails are enabled, and manage patient data through proper data segregation and encryption practices as described in HMCP.
• Troubleshooting: If the server fails to start, verify Python version, environment variable names, port availability, and that dependencies are installed. Check server logs for startup errors related to authentication or transport configuration.