Atlas Agents is a chapter-by-chapter Python example repository for learning how AI agents are built, extended, guarded, evaluated, and deployed. It starts with a raw ReAct loop and ends with a capstone autonomous engineering assistant built around LangGraph, Claude, tool execution, review, and human approval.
This tutorial edition is based on the original repository, agulli/atlas-agents. This version adds extensive tutorial documentation and an interactive HTML course hub for learners who want a guided path through the agent development journey.
Open the learning hub:
The hub turns the repository into an end-to-end course with module navigation, progress tracking, search and filters, lab commands, knowledge checks, and links into the relevant docs and source files.
Start with the documentation hub:
- Interactive course hub
- Docs index
- What this repo is
- Quickstart
- Onboarding guide
- Chapter and command reference
Recommended reading order:
- Interactive course hub
- Onboarding
- Quickstart
- Key concepts
- Agent patterns by chapter
- Skills, tools, and safety
- System design
- Chapter and command reference
- Common issues
- Create a virtual environment.
python -m venv .venv
.\.venv\Scripts\Activate.ps1- Install the shared dependency set.
python -m pip install -r requirements.txt- Copy the environment template and add at least one model provider key.
Copy-Item .env.example .env- Run the first Atlas agent from the repository root.
python ch01_react_from_scratch\atlas_v01.py "What is the Model Context Protocol?"The docs were written from the current repository contents. There is no TOC.md, no package manifest, no test suite at the root, and no existing docs/ directory before this documentation pass.
python -m compileall -q . passes after removing trailing literal \n artifacts from four online examples. See common issues for remaining setup caveats.