Home Introduction: Why Build from Scratch? Phase 1: Build a Working Coding Agent 1. Agent Loop — Core Loop 2. Tool System 3. System Prompt Engineering 4. CLI & Sessions 5. Streaming & Dual Backend 6. Permissions & Security 7. Context Management Phase 2: Advanced Capabilities 8. Memory System 9. Skills System 10. Plan Mode 11. Multi-Agent Architecture 12. MCP Integration Summary 13. Architecture Comparison & Next Steps 14. Testing Guide how-claude-code-works ↗