10. Context

The context window is Claude's "working memory" during a session — everything the model can "see" when generating a response. In current Claude models, it is 200,000 tokens (~500 pages of text). However, in practice, the usable space is considerably smaller.

Why isn't the real limit 200K?

Part of the 200K tokens is consumed by fixed elements that Claude Code loads automatically in every interaction:

Component	Approx. consumption	What it is
Auto-compact buffer	~33K tokens (16.5%)	Fixed reserve for summarization and response generation
System tools	~17K tokens (8.4%)	Internal tool definitions (Read, Edit, Bash, etc.)
Memory files	~7K tokens (3.7%)	Project CLAUDE.md + global + auto-memory
System prompt	~3K tokens (1.3%)	Claude Code internal instructions
Plugins, skills, and agents	~2K+ tokens	Each installed plugin/skill adds overhead

Adding it all up, ~60K tokens (~30%) are consumed before you even type your first prompt. That's why auto-compact triggers when usage reaches ~83% — not 100%. In practice, the actual space for your conversation is approximately ~160-165K tokens.