Iβve been diving into the world of AI Agents, inspired by tools like Claude Code and Dexter. This week, I decided to build my own: Marquee CLIβan autonomous agent that researches movies for you.
The Mission: Build a terminal-based expert that uses a "Plan-Act-Observe" loop to find movies based on my mood, not just a search bar.

β
The Brain: Switched from OpenAI to Google Gemini 2.5 Flash (highly recommend for speed!).
β
The Tools: Built custom Python functions to talk to the TMDb API.
β
The Logic: Used the Agno framework to let the AI "decide" which tool to use.
The Reality Check: Itβs not "production-ready" yet. I hit quota limits, learned the hard way about managing API billing, and spent an hour fighting with .gitignore to keep my keys safe.
My biggest takeaway? Building agents isn't just about the prompt. It's about tool-calling logic and environment security. Itβs 10% AI and 90% engineering.
Huge thanks to the community for the inspiration!
On to the next build. π
check the video below how it works π
github repo :https://lnkd.in/dJeEpuu2
Top comments (0)