Medical Agents AOP Server

Medical Agents AOP Server

A production‑ready template for deploying multiple specialized medical agents as MCP tools using the AOP (Agent Orchestration Protocol) from the `swarms` library. This server exposes agents as callable tools that any MCP‑compatible client can discover and execute.

What You Get

• Multiple domain‑focused medical agents (labs, ICD‑10 mapping, treatment options, drug interactions, imaging triage, clinical note summarization)
• A single AOP server that registers each agent as an MCP tool
• Sensible defaults, safety language, and structured outputs
• Clear examples for customization and MCP client usage

Installation

pip install -r requirements.txt

Quick Start: Run the Server

This repository ships with a ready‑to‑run server in `app.py` that registers six medical agents.

python app.py

By default, the server starts on port `8000` and registers each agent as an MCP tool using the agent’s `agent_name`. You should see logs indicating the server name (`MedicalAgentServer`) and registered tools.

Customizing AOP Settings (Port, Host, Logging, Queue)

In `app.py`, the AOP instance is created like this:

deployer = AOP(
    server_name="MedicalAgentServer",
    description="...",
    port=8000,
    verbose=True,
    log_level="INFO",
)

You can adjust:

• port: Change the listening port (e.g., port=8010)
• host: Bind externally (e.g., host="0.0.0.0")
• verbose and log_level: Increase/decrease logging ("DEBUG", "INFO", "WARNING", "ERROR")
• Queue options for higher throughput:
  • queue_enabled=True
  • max_workers_per_agent=2
  • max_queue_size_per_agent=500
  • processing_timeout=60
  • retry_delay=2.0

Example:

deployer = AOP(
    server_name="MedicalAgentServer",
    port=8010,
    host="0.0.0.0",
    verbose=True,
    log_level="DEBUG",
    queue_enabled=True,
    max_workers_per_agent=2,
    max_queue_size_per_agent=500,
    processing_timeout=60,