MCP协议实现

Aether-MCP: MCP Reference Implementation

Aether-MCP is a reference implementation of the Model Context Protocol (MCP) for clinical-adjacent workflows. It's a small, focused example of how to expose LLM tools over MCP with structured logging and caching.

Status: demo / reference. See "Roadmap" below for what's real vs. simulated.

Features (implemented)

• MCP tool surface — two tools exposed via stdio: query_clinical_database (cache-backed lookup with a simulated result) and orchestrate_validation (multi-step async workflow with simulated latencies).
• Redis caching layer — real Redis-backed cache with TTL, graceful fallback when Redis is unavailable.
• Structured logging — JSON logs for every tool execution.

Roadmap (not yet implemented)

• Real Postgres-backed clinical query (currently returns a canned response).
• Real orchestration work (currently asyncio.sleep() placeholders for latency).
• RBAC + audit-log table.
• Test suite.
• Container image.

🛠 Tech Stack

• Language: Python 3.11+
• Protocol: Model Context Protocol (MCP)
• Framework: FastAPI, Pydantic v2
• Database: PostgreSQL (State), Redis (Caching)
• Deployment: Docker, GitHub Actions (CI/CD)

📦 Installation & Setup

1. Clone the repository:

   复制

   git clone https://github.com/pranavkrishnadanda/Aether-MCP.git
   cd Aether-MCP
   

2. Install dependencies using Poetry:

   复制

   poetry install
   

3. Configure Environment:

   复制

   cp .env.example .env
   # Add your database credentials
   

4. Run the MCP Server:

   复制

   poetry run python -m aether_mcp.server
   

🤖 Usage with Claude / IDEs

Add the following to your MCP configuration (e.g., claude_desktop_config.json):

{
  "mcpServers": {
    "aether-mcp": {
      "command": "python",
      "args": ["-m", "aether_mcp.server"],
      "cwd": "/path/to/Aether-MCP"
    }
  }
}

Author

Pranav Krishna Danda LinkedIn | GitHub