Welcome and setup
Welcome to Build Your Own Claude Code! I'm Param, and in this workshop we'll reverse-engineer how Claude Code works and build our own from scratch. Each phase adds one real capability. By the end, you'll have a production-grade AI coding agent and understand every line.
Here's our journey across 8 learning modules (12 workshop modules): 1. The Agent Loop, the 17-line core that powers everything 2. Tool Design with safe, composable tools and path sandboxing 3. Planning with structured constraints that keep the agent on track 4. Subagents and Skills for context isolation and knowledge externalization 5. Memory Management with three-layer compression for unlimited conversations 6. Parallelism for background execution and persistent teammates 7. Team Coordination with message buses, shutdown protocols, and autonomous agents 8. Worktree Isolation using git worktrees for safe parallel execution Each module builds on the previous one. The code is cumulative.
Workshop roadmap: from loop to production agent
Each module adds one capability. The complexity builds progressively.
You should have completed the Building AI Agents course (or equivalent). You need Python knowledge, a free LLM API key, and a terminal. We use LiteLLM so you can use any provider, and Google AI Studio has a free tier.
# Clone the workshop repository
git clone https://github.com/learnwithparam/learn-to-build-claude-code.git learn-to-build-claude-code
cd learn-to-build-claude-code
# Create .env file and add your API key
cp .env.example .env
# Edit .env with your API key
# Install dependencies and run tests
make setup
make testClone the repo, configure your API key, and verify the setup.
Validation checklist: Environment setup checklist
Loading practice…