README.md

Claude Code Analysis Index

This index page groups the English reports by topic so readers can quickly jump into the parts of Claude Code they care about most: architecture, prompt engineering, telemetry, permissions, MCP, context management, persistence, and hidden internal behavior.

Start Here

If you are new to this repository, begin with:

1. The Agent Loop
2. Tool System Architecture
3. Permission & Security Model
4. Architecture Overview

These three reports explain how Claude Code behaves as a production AI coding agent at the systems level.

Reports by Theme

Architecture and Core Runtime

• 13-claude-code-architecture-overview.md
• 06-agent-loop-deep-dive.md
• 07-tool-system-architecture.md
• 14-bash-tool-security-and-execution.md
• 11-context-window-management.md
• 12-state-management.md
• 16-transcripts-compaction-and-session-resume.md

Prompting, Memory, and Model Behavior

• 09-system-prompt-engineering.md
• 15-memory-and-instruction-system.md
• 03-undercover-mode.md
• 05-future-roadmap.md

Security, Permissions, and Control Surfaces

• 01-telemetry-and-privacy.md
• 04-remote-control-and-killswitches.md
• 08-permission-security-model.md

Integration and Extensibility

• 10-mcp-integration.md
• 02-hidden-features-and-codenames.md

What These Reports Cover

Across the full set of reports, this repository covers:

• Claude Code architecture
• Claude Code source analysis
• Claude Code agent loop internals
• Claude Code system prompts
• Claude Code telemetry and privacy
• Claude Code permission model
• Claude Code MCP integration
• Claude Code context compression
• Claude Code state management
• Claude Code Bash tool
• Claude Code transcript storage
• Claude Code session resume
• Claude Code memory system
• Claude Code hidden features
• Claude Code internal codenames
• Claude Code remote control mechanisms

Reading Paths

If you care about how Claude Code works internally

• 06-agent-loop-deep-dive.md
• 07-tool-system-architecture.md
• 14-bash-tool-security-and-execution.md
• 12-state-management.md
• 16-transcripts-compaction-and-session-resume.md

If you care about prompts, memory, and model behavior

• 09-system-prompt-engineering.md
• 15-memory-and-instruction-system.md
• 11-context-window-management.md
• 05-future-roadmap.md

If you care about infrastructure and networking

• 17-bridge-system-and-remote-sessions.md
• 18-react-ink-terminal-ui.md
• 19-streaming-and-transport-layers.md
• 20-slash-commands-and-cost-tracking.md
• 10-mcp-integration.md

If you care about security and control

• 01-telemetry-and-privacy.md
• 04-remote-control-and-killswitches.md
• 08-permission-security-model.md

If you care about hidden or surprising behavior

• 02-hidden-features-and-codenames.md
• 03-undercover-mode.md
• 05-future-roadmap.md

Keywords

Claude Code, Anthropic, source analysis, reverse engineering, AI coding agent, agent loop, system prompt, telemetry, privacy, permissions, MCP, Model Context Protocol, context window, persistence, hidden features, prompt engineering, tool system, security research