What is simple-agent?

Simple Agent is a tutorial project that shows you how to build a basic AI agent, software that can reason through a problem, pick the right tool to solve part of it, check its own work, and keep going until it has an answer for you. Instead of just chatting, the agent actively decides what to do next at each step. The project uses LlamaIndex, a framework for building AI applications, and runs on a model called Mixtral through a tool called Ollama, which lets you run AI models locally on your own machine. The agent is given a small set of tools, in this case, two plain Python functions that calculate factorials and check whether a number is prime. When you ask it a question, it reasons about which tool to use, calls it, looks at the result, and then decides whether the task is done or if it needs to take another step. It loops like this until it can return a final answer. The project walks through the whole process in a Jupyter Notebook, so you can see each piece working step by step. This is aimed at developers or learners who want a concrete, minimal example of how agentic AI works. If you have heard the term "AI agent" but want to actually see one built from scratch, with real tools, a real model, and visible reasoning, this gives you that in a handful of code cells. You could swap in your own tools and ask the agent to use them instead. One notable choice is that it uses a locally hosted model rather than a cloud API like OpenAI's. That means you do not need an API key or an internet connection to the model, but you do need a machine capable of running Mixtral, which is a fairly large model. The project also keeps things deliberately bare-bones, focusing on the core loop of reasoning and tool use rather than building a polished application.