Inside Claude Code
Inside Claude Code: What Happens When You Type 'claude'
How Claude Code boots in under 200ms using parallel prefetch, lazy loading, and startup optimization. Part 1 of a 10-part series reverse-engineering Anthropic's AI coding tool.
Inside Claude Code: How an AI Agent Thinks in a Loop
The recursive query loop at the heart of Claude Code — how it manages multi-turn reasoning, tool execution, and streaming responses. Part 2 of 10.
Inside Claude Code: Giving AI Hands with the Tool System
How Claude Code's tool contract, factory pattern, and registry pipeline let AI interact with the real world safely. Part 3 of 10.
Inside Claude Code: Why Your AI Should Talk While It Thinks
Claude Code's streaming-first API design — progressive token rendering, concurrent tool execution, and why batch responses are dead. Part 4 of 10.
Inside Claude Code: Permissions in AI Agents
Multi-layered authorization — permission modes, deny rules, ML classifiers, and why no single check is enough. Part 5 of 10.
Inside Claude Code: Plugging Into Everything with MCP
How Claude Code connects to external tools through the Model Context Protocol — and why having four extension systems is three too many. Part 6 of 10.
Inside Claude Code: Remembering What Matters
Context window management, compaction via forked agents, session memory extraction, and the art of forgetting gracefully. Part 7 of 10.
Inside Claude Code: React in Your Terminal
How Claude Code uses React and Ink to bring component-based UI to the terminal — 140+ components powered by familiar patterns. Part 8 of 10.
Inside Claude Code: When One Agent Isn't Enough
Multi-agent coordination — how the coordinator pattern orchestrates parallel AI workers and why coordination is harder than it looks. Part 9 of 10.
Inside Claude Code: Building Something Better
Everything I learned from reverse-engineering Claude Code, distilled into a blueprint for building your own AI coding harness. Part 10 of 10.