Claude Code Unpacked

Claude Code Unpacked is an unofficial, interactive reference site that reverse-engineers and visualizes the internals of Claude Code — Anthropic's AI coding agent. It walks through the full agent loop step by step, catalogs every built-in tool and slash command, and surfaces hidden or feature-flagged capabilities not yet publicly documented. The analysis is based on publicly available source code and is curated with AI assistance.

• Agent Loop Visualization — Step-by-step breakdown of what happens from user keypress to rendered response, covering input, history, system prompt, API call, token handling, tool dispatch, and hooks.
• Tool System Catalog — All 49+ built-in tools organized by category (File Operations, Execution, Search & Fetch, Agents & Tasks, Planning, MCP, System, Experimental), each with source file references.
• Command Catalog — 88+ slash commands grouped into Setup & Config, Daily Workflow, Code Review & Git, Debugging & Diagnostics, and Advanced & Experimental sections.
• Architecture Explorer — Clickable source tree covering tools, components, commands, services, hooks, and more across 1,759+ files and 481K+ lines of code.
• Hidden Features — Documents unreleased or feature-flagged capabilities such as Buddy (terminal virtual pet), Kairos (persistent memory mode), UltraPlan (long Opus planning sessions), Coordinator Mode (parallel multi-agent worktrees), Bridge (remote session control), Daemon Mode, UDS Inbox, and Auto-Dream.
• DeepWiki Integration — Each hidden feature links directly to DeepWiki for deeper AI-assisted exploration of the underlying source.