In this project, I build an AI agent that doesn’t just “scrape a page” the old-school way—it searches, extracts, and then turns the output into a clean Excel sheet automatically. I’m using Firecrawl API for smart web scraping (so I can pull structured content like product title, image, price, and links) and Composio tools to automate the downstream work like generating spreadsheets. The goal is simple: reduce the glue code and make the agent actually useful end-to-end. I walk through how I connect Firecrawl + Composio with Python, and then I build two practical agent workflows: (1) scraping gaming PC details (image, description, price, link) and (2) a job scraping bot that outputs everything into Excel. If you’re learning GenAI system design, this is a great example of tool-using agents: you define the task, give the agent reliable tools, and make the output deterministic enough to ship. Before you start, I also point you to the prerequisites on AI agents and agentic design patterns, because the tooling is easy—the architecture is what makes it scalable.