Build a ReAct Agent

What you'll be able to do

✓Build the ReAct loop from scratch in pure Python — reason, act, observe, repeat
✓Define a tool set of callable functions with names and descriptions the planner can dispatch
✓Write the planner contract — the Thought / Action / Action-Input format the LLM must emit
✓Parse a planner step and route it to the right tool, feeding each observation back in
✓Harden the loop against sloppy output — the format-and-parsing fix that lifts reliability from ~20% to ~60%
✓Ship a reusable ReActAgent class and swap the scripted planner for a real LLM in one line

Build the reason-act-observe engine behind tool-using AI agents — the loop that lets ChatGPT, Claude, and Perplexity call a calculator, search the web, or run code instead of guessing from memory.

⌁ The core loop inside every tool-using agent — OpenAI function-calling agents, Anthropic's Claude tool use, LangChain and LlamaIndex agents, Perplexity's answer engine, and AutoGPT-style task runners.

Start this internship

Create an account to unlock the 9 sections, the workbench, and AskThili.

Begin