Claude Code
Claude Code is Anthropic's official CLI for Claude. It runs in your terminal, reads your codebase, executes commands, and can handle multi-step programming tasks autonomously.
CLAUDE.md
Create a CLAUDE.md file at the root of your project to give Claude Code persistent context: coding conventions, project structure, commands to run, things to avoid.
Managing Context
Claude Code's output quality degrades as the context window fills up. Understanding and managing context usage is critical for long sessions.
Permissions
Claude Code uses a layered permission system to control which tools can run and which files can be accessed.
CLI Flags
Session management
Debugging
When things are not working as expected:
Tips
- Be specific: tell Claude exactly what you want and what constraints to respect
Workflows & Orchestration
Claude Code's commands, subagents, and skills can be composed into structured workflows that go far beyond single-prompt interactions. This page covers orchestration patterns for organizing complex, multi-step tasks.
Skills (Custom Slash Commands)
Skills are reusable prompt templates that let you encode recurring tasks, team conventions, and workflows into short, invocable commands. Instead of re-typing the same instructions every session, you define them once and invoke them with a / prefix.
Plugins
Plugins are installable packages that extend Claude Code with additional skills, agents, hooks, and tool integrations. Where a skill is a single reusable prompt you keep in your project, a plugin bundles multiple components together into a versioned, shareable unit that can be discovered and installed from a marketplace.
MCP Servers
MCP (Model Context Protocol) is an open standard that lets AI coding tools connect to external services: file systems, databases, APIs, browsers, and more. Each MCP server exposes a set of tools the AI can call during a session, extending what it can do without you having to copy-paste content or context manually.
Subagents
Subagents are the primary way Claude Code delegates work without ballooning your main conversation. The main agent spawns a subagent, gives it a focused job, and gets back a clean summary, keeping your context window tidy and your token bill reasonable.
Agent Teams
Agent Teams is an experimental Claude Code feature that lets a lead agent spawn and coordinate multiple specialized teammate agents working in parallel. Teammates communicate through a shared task list and direct messaging, and can optionally be displayed side-by-side in split panes (via tmux or iTerm2).
Memory
When you use an AI coding tool, every new session starts with a blank slate. The model has no recollection of your project's conventions, past decisions, or the patterns your team follows. Memory solves this problem by giving agentic coding tools persistent context that carries over between sessions.