Build Your First AI Agent from Scratch in Python (No Frameworks)

Companion repository for the Code with Antonio video course · Watch on YouTube.

Build Your First AI Agent from Scratch in Python, with no frameworks: tools, the agent loop, memory, human approval and step limits.

What you'll build

A notes agent for your terminal: it tells the time, lists, reads and writes the files in a workspace folder, remembers the conversation, asks before it saves anything and knows when to stop.

Lessons

#	Lesson	Video	Folder
1	What is an AI agent, really?	0:00	lessons/01-what-is-an-ai-agent-really
2	Set up the project with uv	2:59	lessons/02-set-up-the-project-with-uv
3	Give the model its first tool	6:53	lessons/03-give-the-model-its-first-tool
4	The agent loop	9:41	lessons/04-the-agent-loop
5	Memory: a chat that remembers	11:58	lessons/05-memory-a-chat-that-remembers
6	Tools that act — safely	14:28	lessons/06-tools-that-act-safely
7	Knowing when to stop	17:18	lessons/07-knowing-when-to-stop

Requirements

• OpenAI Python SDK 3.22.1 · Responses API · gpt-6-luna · uv 0.12.21 · Python 3.12 · Ubuntu (Linux)
• Python 3.12+ and uv

Run the final project

cd app
uv sync
cp .env.example .env   # then add your own keys (e.g. OPENAI_API_KEY)
uv run --env-file .env agent

Repository structure

README.md          this file
lessons/           one folder per lesson: goal, timestamp, commands and files exactly as in the video
app/               the final working project
assets/            thumbnail

License

MIT — use it for anything, credit appreciated.

---

Generated on 2026-10-01 from the course scripts. Code with Antonio · AI tutor.