Learn Claude Code

Learn Claude Code is an educational project that walks you through building a minimal, functional agent harness inspired by Claude Code—entirely in Bash. It strips away complexity to reveal the core mechanics of agentic loops: tool calling, environment sensing, and self-correction. Designed for developers who want to understand, not just use, AI agents, this hands-on tutorial covers everything from zero to a working shell-based agent. The project is structured as a step-by-step guide, with each stage introducing a new capability: reading files, executing commands, interpreting LLM responses, and chaining actions. By the end, you'll have a functional “nano agent” that can perform tasks like code editing and shell operations, all orchestrated through Bash scripts. It's a pure learning tool—not production-grade, but a brilliant way to demystify agent architectures. Who is this for? Intermediate developers curious about AI agent internals, Bash enthusiasts, and anyone who finds existing agent frameworks too abstract. It's particularly valuable if you prefer understanding over black-box usage. The project assumes basic familiarity with the command line and a willingness to read and modify shell scripts. What sets it apart is its radical simplicity: no dependencies beyond Bash and a shell environment. It's the anti-framework—a single, transparent codebase that reveals how agents really work. Perfect for tinkerers, educators, and anyone who learns best by building from first principles.