Case study
Boot.dev Agent Project
A guided Boot.dev project that walks through the basics of LLM tool use. The agent exposes a small set of file and calculator functions to the Gemini API and runs a multi-turn loop until the model returns a final answer.
The brief
Challenge
Build an end-to-end LLM agent loop from scratch without an agent framework.
Approach
What we made
Followed the Boot.dev guided project: defined tool schemas, parsed model tool calls, executed each tool, and fed results back into the next model turn until completion.
- Implements a basic tool-calling loop against the Gemini API.
- Exposes file inspection and calculator helpers as agent tools.
- Built with uv for reproducible Python environments.
Outcome
Results
A small but complete agent showing the core tool-use loop pattern that scales up to larger production agents.
PythonGemini APIuv
Gallery
Visual snapshots
Click any image to expand.