Claude Code: The Complete Guide

Claude Code is Anthropic’s agentic coding tool — an AI assistant that lives in your terminal (and inside your editor) and can actually read your codebase, write and edit files, run commands and tests, and carry a multi-step task through to completion. Instead of copy-pasting snippets from a chat window, you delegate real work and review the results.

This guide explains what Claude Code is, how it works, what it can do, what it costs, and how it compares to tools like Cursor and GitHub Copilot — with links to deeper guides for each topic.

What is Claude Code?

Claude Code is a command-line tool, powered by Anthropic’s Claude models, that turns natural-language instructions into concrete changes in your project. You describe a goal — “add pagination to the posts API and write tests” — and Claude Code explores the relevant files, proposes a plan, makes the edits across multiple files, runs your test suite, and iterates until the task is done.

It runs where developers already work: the terminal, with official extensions for VS Code and JetBrains IDEs, plus desktop and web access. Because it’s terminal-native, it also slots neatly into scripts, CI pipelines, and remote servers.

What can Claude Code do?

• Understand your whole codebase — it searches and reads files on demand, so you don’t have to paste context manually.
• Make multi-file edits — refactors, new features, and bug fixes that span many files in one task.
• Run commands and tests — it can execute your build, run tests, read the output, and fix what broke.
• Use git — create branches, stage changes, write commit messages, and open pull requests.
• Connect to your tools via MCP — databases, issue trackers, documentation, and more through the Model Context Protocol.
• Work autonomously but safely — it asks before running sensitive commands, and a plan mode lets you approve an approach before any files change.

How Claude Code works

Under the hood, Claude Code gives a Claude model a set of tools (read file, edit file, run command, search, git, and any MCP tools you add) and lets it act in a loop: think, take an action, observe the result, and continue. You stay in control — reviewing diffs, approving commands, and steering with follow-up instructions.

A small CLAUDE.md file in your project acts as persistent memory: coding conventions, commands, and project context that Claude Code reads at the start of every session. Slash commands, subagents, and hooks let you customise and automate the workflow further — see Claude Code best practices.

Which Claude model powers it?

Claude Code’s default model depends on your plan: Claude Sonnet 5 on Pro, Team Standard, and Enterprise seats, and Claude Opus 5 on Max and Team Premium. You can switch models per task with the /model command, or set ANTHROPIC_DEFAULT_MODEL to choose what new sessions start on. Use a flagship model like Claude Opus 5 — Anthropic’s recommended default since July 24, 2026 — for hard, multi-step work, keep Sonnet 5 for everyday sessions to save on usage, and reach for Claude Fable 5.1 (available in Claude Code from its September 1, 2026 launch) for multi-hour agentic runs where Opus 5 at high effort still falls short. Not sure which to pick? Try the Claude model selector.

Who is Claude Code for?

It’s built for software developers — from solo builders to large engineering teams — who want to move faster on real tasks: shipping features, fixing bugs, writing tests, migrating code, and understanding unfamiliar codebases. If you’re new to using Claude for development, start with Claude AI for coding and our Claude coding prompts.

How much does Claude Code cost?

Claude Code is included with Claude Pro and Claude Max subscriptions (subject to usage limits), or you can pay per token with an Anthropic API key. Full breakdown in Claude Code pricing & usage limits.

How does it compare to other AI coding tools?

The biggest difference is autonomy: Claude Code is an agent that completes tasks, where editor tools focus on autocomplete and in-line chat. See the head-to-head guides:

• Claude Code vs the Claude chat app
• Claude Code vs Cursor
• Claude Code vs GitHub Copilot

How to get started

Install it with one command, sign in, and point it at a project. Our getting-started guide walks through installation, authentication, your first session, and IDE setup step by step — then keep the commands cheat sheet handy for the slash commands and shortcuts you’ll use daily.

Recent Claude Code updates (August 2026)

Claude Code ships weekly. The changes most likely to affect how you work, from the official release digests for July and August 2026:

• Auto mode is now the default permission mode for new sessions on Pro, Max, and Team plans (since August 14, 2026): a classifier approves safe actions and blocks risky ones, replacing most permission prompts.
• /design (research preview) brings Claude Design’s artboard workflow into the CLI and Desktop — Claude drafts editable UI artboards and implements the one you pick.
• Fork mode on by default, plus /fork to copy a conversation into a background session, and cross-session messaging — type @ to hand a finding from one session to another.
• Auto-continue after a usage limit on Desktop: check the option on the limit card and the interrupted turn retries when limits reset.
• Self-hosted environments for cloud sessions (public beta on Team and Enterprise), and GitLab merge-request support with --worktree.
• Claude Opus 5 is the default Opus model, with fast mode at $10/$50 per MTok; ANTHROPIC_DEFAULT_MODEL sets your starting model.
• Concise output style leads with the result and skips preamble; remote-control device cards let you start a session on any of your machines from your phone.
• Claude in Chrome is generally available, Desktop has an in-app browser and an iOS Simulator pane (beta), the Claude Security plugin runs multi-agent vulnerability scans, and /code-review runs as a background subagent.

Frequently Asked Questions

Explore the Claude Code guide