Everything Claude Code: The Complete 2026 Guide

Claude Code is Anthropic's agentic coding tool: it reads your codebase, edits files, runs commands, and integrates with your development tools, available in the terminal, your IDE, a desktop app, and the browser. This guide is the complete hub for everything Claude Code, from installing it in one line to wiring up MCP servers, skills, subagents, agent teams, hooks, and settings, with every command verified against the official Claude Code documentation as of July 2026. If you want one page that maps the whole system and points to a deeper guide for each piece, this is it.

What Is Claude Code?

Claude Code is an AI-powered coding assistant that understands your entire codebase and can work across multiple files and tools to build features, fix bugs, and automate development tasks. Unlike autocomplete tools that suggest the next line, it is agentic: you describe an outcome in plain language, and it plans the approach, edits code across files, runs the tests, and verifies the result. It works directly with git too, staging changes, writing commit messages, creating branches, and opening pull requests.

The same engine runs on every surface, so your CLAUDE.md files, settings, and MCP servers carry across all of them. The terminal CLI is the full-featured surface; there are also extensions for VS Code and JetBrains IDEs, a desktop app, and a browser version at claude.ai/code for long-running tasks. Our Claude Code guide compares those surfaces and where each fits, and Claude Code vs Cursor covers how it stacks up against IDE-first tools.

Install and First Run

The recommended install is the one-line native installer, which downloads a self-contained binary that needs no Node.js and auto-updates in the background. Pick the command for your shell:

# macOS, Linux, WSL
curl -fsSL https://claude.ai/install.sh | bash

# Windows PowerShell
irm https://claude.ai/install.ps1 | iex

Homebrew (brew install --cask claude-code), WinGet (winget install Anthropic.ClaudeCode), and npm (npm install -g @anthropic-ai/claude-code, which needs Node.js 22+) are alternatives. After installing, run claude in a project directory and log in through the browser prompt, then verify with claude --version and diagnose issues with claude doctor. The full platform-by-platform walkthrough, including the Windows and WSL differences, lives in our how to install Claude Code guide.

Pricing Tiers

Claude Code is included with paid Claude subscriptions and is not available on the free Claude.ai plan. You can also authenticate with an Anthropic Console account and pay per token at API rates, which suits automation and CI. Prices below are current as of July 2026 from claude.com/pricing.

On subscription plans, usage counts against your plan limits rather than a per-token bill. Across enterprise deployments Anthropic reports an average of about $13 per developer per active day and $150 to $250 per developer per month on API billing, so a Pro or Max subscription is usually cheaper for steady individual use. Our is Claude Code free guide covers what you can and cannot do without paying.

Core Workflows

The core loop is conversational: you type a request, Claude Code explores the codebase, proposes changes, and applies them once you approve. A few workflow features do most of the heavy lifting, and getting fluent with them is the difference between a novelty and a daily tool.

• Plan mode. Press Shift+Tab to enter plan mode before a complex task. Claude explores the code and proposes an approach for your approval in read-only mode, which prevents expensive re-work when the first direction is wrong.
• Slash commands. Built-ins like /clear (reset context between tasks), /model (switch models mid-session), /context (see what is filling the window), /usage (token usage), and /rewind (restore to a checkpoint) run the session. See our Claude Code commands reference.
• CLI scripting. Claude Code follows the Unix philosophy, so you can pipe into it and run it in CI: tail -200 app.log | claude -p "flag any anomalies" or git diff main --name-only | claude -p "review these files for security issues".
• Memory. A CLAUDE.md file at your project root is read at the start of every session; use it for coding standards and architecture notes. Claude also builds auto memory as it works, saving build commands and debugging insights across sessions.

For habits that hold up over months, our Claude Code best practices guide is the companion piece. You can also customize the terminal display with a status line that shows context usage continuously.

MCP Servers

The Model Context Protocol (MCP) is an open standard for connecting Claude Code to external data sources and tools, so it can read Jira issues, query a Postgres database, or pull data from Slack directly instead of you pasting it in. You add a server with the claude mcp add command, choosing a scope (local, project, or user) and a transport (stdio for local processes, or HTTP for remote servers).

# Local stdio server
claude mcp add context7 -- npx -y @upstash/context7-mcp

# Remote HTTP server
claude mcp add --transport http github https://api.githubcopilot.com/mcp/

MCP tool definitions are deferred by default, so only tool names enter context until Claude actually uses one, which keeps the window lean. Browse our MCP server directory of 13,870 servers across 14 categories, see the best MCP servers of 2026, or read the deeper Claude Code MCP walkthrough for scopes and troubleshooting.

Skills

Skills extend what Claude can do by packaging repeatable workflows into a SKILL.md file with YAML frontmatter, loaded on demand only when relevant so long reference material costs almost nothing until you need it. They follow the open Agent Skills standard and live in .claude/skills/<name>/SKILL.md for a project or ~/.claude/skills/ for all projects. Claude invokes a skill automatically when the task matches its description, or you can call it directly with /skill-name.

Custom slash commands have merged into skills: a file at .claude/commands/deploy.md and a skill at .claude/skills/deploy/SKILL.md both create /deploy. Moving long procedures out of CLAUDE.md and into skills is one of the best token savers, because CLAUDE.md loads every session while skills load only when used. Browse the skills directory to install ready-made ones, and see where else to find them in our best places to find Claude Code skills roundup.

Subagents and Agent Teams

Subagents are specialized assistants that run in their own context window with a focused system prompt and limited tools, so a side task like research or test-running does not flood your main conversation. They are Markdown files with YAML frontmatter in .claude/agents/ (project) or ~/.claude/agents/ (user); only name and description are required, and you can set tools and model (for example model: haiku to route cheap tasks to a faster model).

---
name: code-reviewer
description: Reviews code for quality and best practices
tools: Read, Glob, Grep
model: sonnet
---
You are a senior code reviewer. Focus on correctness, security, and clarity.

Claude delegates to a subagent automatically when a task matches its description. For picking which agents to define first, see our ranked best Claude Code subagents guide. When workers need to talk to each other rather than just report back, use experimental agent teams, where a lead session coordinates teammates that share a task list and message directly. Our Claude agent teams piece covers structuring that work without chaos.

Hooks and Settings

Hooks run shell commands automatically at lifecycle events, so you can auto-format after every file edit, run lint before a commit, or block risky commands. They are configured in your settings under a hooks block, keyed by event. Common events include PreToolUse and PostToolUse (around every tool call), UserPromptSubmit, SessionStart and SessionEnd, Stop, and SubagentStop.

{
  "hooks": {
    "PostToolUse": [
      {
        "matcher": "Edit|Write",
        "hooks": [{ "type": "command", "command": "~/.claude/hooks/lint.sh" }]
      }
    ]
  }
}

Settings live in layered JSON files that override each other in a set order: managed (enterprise) settings win, then command-line arguments, then project-local .claude/settings.local.json, then shared .claude/settings.json, then user ~/.claude/settings.json. Permission rules are the exception, merging across scopes rather than overriding. The full event list and handler types are in our Claude Code hooks guide.

Plugins

Plugins bundle skills, agents, hooks, and MCP servers into one shareable, versioned package, which is how teams distribute a whole Claude Code setup at once. You add a marketplace and install a plugin from inside Claude Code:

/plugin marketplace add anthropics/claude-plugins-community
/plugin install some-plugin@claude-community

Anthropic maintains two public marketplaces: claude-plugins-official (curated) and claude-community (third-party submissions after review). Plugin skills are namespaced, so a skill in a plugin named my-tool is invoked as /my-tool:skill-name to avoid conflicts. Our Claude plugins guide and best Claude Code plugins of 2026 cover installation and the ones worth adding. Note the popular community collection often searched as "everything claude code" (the ECC repo) is one such bundle you can cherry-pick from rather than adopt wholesale.

Tips That Save Tokens

Token cost scales with context size, so the highest-leverage habits are the ones that keep the window small. A few that consistently pay off:

• Clear between tasks. Run /clear when switching to unrelated work so stale context stops costing tokens on every message.
• Right-size the model. Sonnet handles most coding well and costs less than Opus; reserve Opus for hard architectural reasoning and use /model to switch mid-session.
• Prefer CLI tools over MCP where possible. Tools like gh and aws add no per-tool listing to context, while MCP servers do; run /context to see what is consuming space.
• Delegate verbose work. Push test runs and log parsing to subagents so the output stays in their context and only a summary returns to yours.
• Write specific prompts. "Add input validation to the login function in auth.ts" reads far fewer files than "improve this codebase".

You can browse standalone free utilities and supporting guides in our tools hub to round out a workflow without adding context overhead.

Limitations and Tradeoffs

Claude Code is powerful, but it is not a set-and-forget autopilot. It can head down the wrong path on vague requests, and every MCP server, skill, and always-on CLAUDE.md line consumes context on each turn, which is why the docs recommend keeping CLAUDE.md under 200 lines and disabling MCP servers you are not using. Agent teams multiply that cost, using roughly 7x the tokens of a single session in plan mode, so they are worth it for parallel research and review but wasteful for routine tasks.

It also requires a paid plan and an internet connection, so there is no free or offline path. Treat generated code the way you would a capable but junior teammate's: review the diffs, run the tests, and keep secrets out of prompts (a secret-detection hook is a cheap safety net). Used with that discipline, it earns its place; used blindly, it produces plausible code you then have to debug.

Related Guides

• How to Install Claude Code on Mac, Windows, and Linux
• Claude Code MCP: How to Add and Manage MCP Servers
• Best Claude Code Subagents in 2026
• Claude Code Agent Teams: Setup and Orchestration