Ways to extend Claude Code

Extensions plug into different parts of the agentic loop. Each solves a different problem:

Feature	What it does	When to use it
CLAUDE.md	Persistent context loaded every conversation	Project conventions, “always do X” rules
Skill	Instructions, knowledge, workflows Claude can use	Reusable content, repeatable tasks
Subagent	Isolated execution context, returns summaries	Context isolation, parallel tasks
MCP	Connect to external services	External data or actions
Hook	Script/prompt triggered by lifecycle events	Automation that must run on every matching event
Plugin	Packages the above into one installable unit	Reuse a setup across repos, distribute to others

Context cost by feature — CLAUDE.md loads in full every request; skills, MCP, subagents, and hooks defer.

Build your setup over time. You don't configure everything up front — each feature has a recognizable trigger: Claude gets a convention wrong twice → CLAUDE.md. You keep typing the same prompt → skill. You keep copying data from a browser tab → MCP. A side task floods your conversation → subagent. Something must happen every time → hook. A second repo needs the same setup → plugin.

Guardrails go in hooks

"Never edit .env" in CLAUDE.md is a request, not a guarantee. A PreToolUse hook that blocks the edit is enforcement.

Check your understanding

Question 1A side task keeps flooding your conversation with output you won't reference again. Add a…

AHookBSubagentCCLAUDE.md ruleDTheme

Question 2A rule must hold every single time, no exceptions. Use a…

ASkillBCLAUDE.md lineCHook (or permission rule)DOutput style

Question 3Which feature is the packaging layer for the others?

AMCPBPluginsCRulesDWorkflows

Further reading

• Extend Claude Code