This is a teaching project from an OpenAI API agent development course, bundling both a web agent interface and an MCP server in one deployment. You define your tools in tools.py, point it at an OpenAI prompt ID, and get function calling working through either the web UI or Claude Desktop via the MCP endpoint. It's built with uv for fast dependency management and includes sample fine-tuning datasets for supervised, preference, and reinforcement learning workflows. The source is mostly setup instructions and boilerplate, so you're getting a scaffold for learning agent patterns rather than a production service. Useful if you're following the AI Castle course or want a template that demonstrates wrapping OpenAI agents as MCP tools.
본 자료는 (주)에이아이캐슬에서 만든 OpenAI API로 배우는 Agent 개발 첫걸음 (OpenAI API Agent School) 강의 프로젝트 자료입니다.
# uv Install
curl -LsSf https://astral.sh/uv/install.sh | sh
# uv Build
uv sync --frozen && uv cache prune --ci
OPENAI_API_KEY: Agent 앱 또는 파인튜닝할 데이터를 업로드할 때 사용할 OpenAI API 키PROMPT_ID Agent 앱에서 사용할 OpenAI 프롬프트 IDTITLE: Agent앱의 상단 제목PASSWORD: 비밀번호 설정 (비워둘 경우 누구나 접근 가능)
?password=<your-password>와 같이 쿼리스트링으로 전달해야 접근 가능uv run main.py
포트: 환경변수 PORT값이 지정된 경우 이 값을 사용하며, 그렇지 않을 경우 8000을 사용함.
agent 앱 주소: https://localhost:8000/agent
mcp 서버 주소: https://localhost:8000/mcp
KEEPALIVE_URL를 github actions의 환경변수(secrets)에 지정하여 주기적으로 접속하는 cron 작업을 수행할 수 있음.KEEPALIVE_URLhttps://<your-project-name>.onrender.com.env파일에 OPENAI_API_KEY를 등록해야 정상적으로 업로드 가능
fine_tuning_data/supervised/uv run fine_tuning_data/supervised/convert_and_upload.py
fine_tuning_data/preference/uv run fine_tuning_data/preference/convert_and_upload.py
fine_tuning_data/reinforcement/uv run fine_tuning_data/reinforcement/convert_and_upload.py
io.github.ericm1018/skillfm-llm-cost-optimizer-openai-anthropic-usage
io.github.mikerawsonnz/llm-orchestration-agent
io.github.mikerawsonnz/authenticated-llm-agent
labforgedev/copilot-memory-mcp
csoai-org/agent-prompt-injection-firewall-mcp
io.github.mikerawsonnz/authenticated-multi-llm-agent