Claude Code Unpacked

Summary

An interactive deep-dive into Claude Code’s internals - the agent loop, 40+ built-in tools, multi-agent orchestration, slash commands, and unreleased features - all mapped directly from the open source. Essential reference for anyone building on top of or customizing Claude Code.

Key Insight

• Claude Code has 40+ built-in tools organized into clear categories: File Operations (6), Execution (3), Search & Fetch (3), Agents & Tasks (10), Planning (4), MCP (4), System (8), and Experimental (4)
• The agent loop follows a specific path: user input via Ink’s TextInput component (or piped stdin in non-interactive mode) through core processing, tool execution, and back to the UI layer
• The architecture is split into six domains: Tools & Commands, Core Processing, UI Layer, Infrastructure, Support & Utilities, and Personality & UX
• Slash commands are organized into 5 groups: Setup & Config (12), Daily Workflow (24), Code Review & Git (13), Debugging & Diagnostics (23), and Advanced & Experimental (23) - totaling ~95 commands
• The Agents & Tasks category having 10 tools confirms Claude Code has significant multi-agent orchestration built in, useful for understanding how subagent dispatching and task delegation works under the hood
• Source code is browsable per-tool and per-command, making this a living reference rather than static documentation