In this video, I build a custom AI Research Assistant from scratch using OpenAI’s new Agents SDK in Python. I walk you through the core building blocks—Agents (the worker), Handoffs (delegation across steps), and Guardrails (keeping outputs safe/compliant)—and then I stitch them into a practical research workflow you can reuse for your own projects. The goal isn’t “toy agent demos”; it’s system-design thinking you can apply fast. I show how to make the agent generate smarter search queries, pull and summarize information, cite sources, and finally export everything as a clean Markdown research report. Along the way, I highlight how to integrate real automation tools/APIs (like web scraping and tool connectors) so the agent can actually do work instead of just chatting. If you want a repeatable template for agentic research—topic in, structured report out—this project is the exact blueprint, plus I share code/resources so you can extend it into your own RAG + agent stack.