Claude Code Inspect

512K+Lines of TypeScript

46Built-in Tools

5Agent Types

27Bash Security Layers

What You’ll Learn

Agentic Loop Engine ⭐

How a while(true) loop drives the entire AI agent — 7 stages, async generators, streaming tool execution, and 6 stop conditions. Read more →

Multi-Agent System ⭐

From single agent to Fork → Coordinator → Team/Swarm. Complete evolution of a production multi-agent architecture. Read more →

27-Layer Bash Security

The most thorough command security in any AI tool — AST parsing, semantic analysis, and 27 decision checkpoints. Read more →

Reusable Design Patterns ⭐

7 battle-tested patterns extracted from production code: async generator loops, streaming execution, context compression, and more. Read more →

Reading Paths

🚀 Quick Tour (2 hours)

For busy developers

• Architecture Map
• The Agentic Loop
• Agent Evolution
• Lessons for Builders

🔧 Agent Developer (4 hours)

Build your own agent system

• Agentic Loop (all 5 pages)
• Tool System
• Agent System (all 6 pages)
• Design Patterns

🛡️ Security Researcher (3 hours)

Understand AI agent security

• Permission Model
• Bash 27-Layer Security
• Threat Model
• Layered Security Pattern

Explore by Section

📖 Overview

What is Claude Code, architecture map, codebase stats, and how to read this analysis.

🏗️ Foundations

Entry point, configuration system, streaming infrastructure, and error handling.

⭐ Agentic Loop

The core engine — while(true) loop, turn lifecycle, tool dispatch, stop conditions.

🔧 Tool System

46 built-in tools, streaming executor, hook system, and MCP integration.

⭐ Agent System

Single agent → Fork → Coordinator → Team/Swarm. The complete multi-agent architecture.

🧠 Context Engine

System prompt assembly, prompt cache strategy, 4-layer compression, and memory.

🛡️ Permission & Security

Permission model, 27-layer Bash security, rule system, and threat analysis.

⭐ Design Patterns

7 reusable patterns extracted from Claude Code — ready to apply in your own projects.

🔍 Deep Critique

Honest analysis of strengths, weaknesses, trade-offs, and lessons for agent builders.

---

Built with ❤️ for the AI agent developer community.
This is an independent analysis project, not affiliated with Anthropic.