agent-start-project

FastAPI playground API with MCP metadata and a CLI AI agent that uses MCP and RAG to interact with it.

Chosen platform

• ChromaDV as vector database
• OpenAI as example.

Installation

source .venv/bin/activate

pip install -r requirements.txt

Execution REST service

uvicorn api.main:app --reload --host 127.0.0.1 --port 5000

Execution CLI Agent

Illustrating multiple rules folders and multiple MCP servers endpoints

export OPENAI_API_KEY=your_api_key_here
python agent/main.py --rules ./rules ./rules --mcp http://127.0.0.1:5000/mcp http://127.0.0.1:5000/mcp -t ./tasks/example.md ./tasks/example.md