What Is Claude Code? Anthropic's Coding Agent Explained
What is Claude Code? A plain guide to Anthropic's agentic coding tool: what it does, where it runs, what it costs, and every guide you need to master it.
By Claudekit Team Published 7 min read
Claude Code is Anthropic’s agentic coding tool. You describe a task in plain English, and it reads your codebase, edits files across the project, runs commands such as tests and git, and checks its own work. It runs in your terminal, in VS Code and JetBrains, in the Claude desktop app and in the browser, on Claude models.
That is the short answer. The rest of this page explains how it works, where it runs, what it costs, how it compares to other AI coding tools, and links to every in-depth Claude Code guide on Claudekit so you can go from “what is it?” to productive in one sitting. Facts are checked against the official Claude Code docs as of October 2026.
What Claude Code actually does
Anthropic describes Claude Code as “an agentic coding tool that reads your codebase, edits files, runs commands, and integrates with your development tools.” The key word is agentic. A chat assistant answers one message at a time; an agent plans steps, calls tools, looks at the result and keeps going until the job is done or it needs your input.
In practice that means you can ask for things like:
claude "write tests for the auth module, run them, and fix any failures"
claude "commit my changes with a descriptive message"Typical jobs from the official docs:
- Build features and fix bugs. Describe the change or paste an error; Claude Code traces the cause through your files and implements a fix.
- Do the tedious work. Writing tests, fixing lint across a project, resolving merge conflicts, updating dependencies, writing release notes.
- Work with git. Stage changes, write commit messages, create branches and open pull requests.
- Script and automate. Pipe logs into it or run it in CI with
claude -p(see our headless mode guide). - Connect your tools. Read docs, tickets and databases through MCP servers.
You stay in control. Claude Code asks before risky actions, and permission modes let you choose how much it can do without asking.
Claude Code vs the Claude chat app
People often confuse the two, so here is the difference at a glance.
| Claude (chat app) | Claude Code | |
|---|---|---|
| What it is | Conversational assistant at claude.ai and in the apps | Agentic coding tool for software projects |
| Works on | What you paste or upload | Your real repository and file system |
| Runs commands | No (on your machine) | Yes: tests, builds, git, scripts, with your permission |
| Edits files | Produces text and artifacts you copy | Edits files in place, shows diffs |
| Project memory | Projects and memory | CLAUDE.md, auto memory, skills |
| Plan needed | Free tier available | Paid plan, Console (API) account or cloud provider |
If you want to know more about the chat side, read what Claude is. If you are curious which model sits underneath, our Claude models explained page covers Opus, Sonnet and Haiku.
Where Claude Code runs
Every surface connects to the same engine, so your CLAUDE.md files, settings and MCP servers carry over between them.
| Surface | Best for | Setup |
|---|---|---|
| Terminal CLI | Full feature set, scripting, power users | Native installer, Homebrew or WinGet |
| VS Code (and Cursor) | Inline diffs and plan review in the editor | Install the “Claude Code” extension |
| JetBrains IDEs | IntelliJ, PyCharm, WebStorm users | Plugin plus the CLI |
| Desktop app (Code tab) | Visual diff review, parallel sessions | Download the Claude app; paid plan |
| Web at claude.ai/code | Long tasks in the cloud, repos you don’t have locally | Browser or Claude mobile app |
| GitHub Actions / GitLab CI | PR review and issue triage | Workflow file |
| Slack | Turn a bug report into a pull request | Mention @Claude |
Our VS Code and JetBrains guide covers the editor setup, the Claude desktop app guide covers the Code tab, and the GitHub Actions guide covers CI.
How to install it
The official native installer is the recommended route and updates itself in the background:
# macOS, Linux, WSL
curl -fsSL https://claude.ai/install.sh | bash# Windows PowerShell
irm https://claude.ai/install.ps1 | iexThen start it inside a project:
cd your-project
claudeIt asks you to sign in on first run. Homebrew (brew install --cask claude-code) and WinGet (winget install Anthropic.ClaudeCode) also work but do not auto-update. Our step-by-step how to install Claude Code guide covers every OS, PATH problems and WSL.
What Claude Code costs
The software is free; the model usage is not. As of October 2026 you can use Claude Code with:
- a Claude subscription (Pro, Max, Team or Enterprise), where usage counts against your plan limits;
- a Claude Console account, billed per token at API rates;
- a cloud provider such as Amazon Bedrock, Google Cloud or Microsoft Foundry.
The free Claude.ai plan does not include it. Prices and limits change, so see our Claude Code pricing guide for current numbers and the Claude Pro vs Max comparison if you are choosing a plan.
How it works under the hood
Four ideas explain most of Claude Code’s behaviour.
1. The agent loop. Claude reads the request, decides which tool to use (read a file, search, edit, run a command), looks at the result and repeats. You can interrupt with Esc at any point.
2. Context. Everything it reads fills a context window. Managing that well is the biggest skill; our context window guide explains /context, /compact and /clear.
3. Project memory. A CLAUDE.md file in your repo root is read at the start of every session, so Claude learns your conventions once. It can also read an existing AGENTS.md. Start with our CLAUDE.md guide.
4. Extensions. Claude Code is customisable in layers:
| Extension | What it adds | Guide |
|---|---|---|
| Skills | Reusable procedures and slash commands | Skills guide |
| Subagents | Specialists with their own context window | Subagents guide |
| Hooks | Shell commands that run on events, every time | Hooks guide |
| MCP servers | Connections to GitHub, docs, browsers, databases | MCP servers for Claude |
| Plugins | Installable bundles of all of the above | Plugins guide |
Not sure which to reach for? Read skills vs subagents vs hooks. New to the protocol behind MCP? Start with what is MCP.
How Claude Code compares
Claude Code is one of several AI coding agents. The honest summary: it is strongest when you want a deeply customisable agent that works across terminal, IDE, desktop and CI. Editor-first tools suit people who want AI inside a familiar IDE.
| Comparison | Read |
|---|---|
| Claude Code vs Cursor | Full comparison |
| Claude Code vs OpenAI Codex | Full comparison |
| Claude Code vs GitHub Copilot | Full comparison |
| Claude Code vs Gemini CLI | Full comparison |
| Claude Code vs Windsurf | Full comparison |
| Claude vs ChatGPT for coding | Full comparison |
Every Claude Code guide on Claudekit
Use this as your reading path. Each link goes deeper than this overview.
Start here
- How to install Claude Code: Mac, Windows, Linux and WSL.
- Getting started with Claude Code: from install to your first pull request.
- Claude Code tutorial: one real task, start to finish.
- Claude Code FAQ: fifteen common questions answered.
- Claude Code pricing: plans, API costs and limits.
Work well day to day
- Claude Code best practices: twenty habits that make a difference.
- Slash commands cheatsheet: the commands worth memorising.
- Context management: keep long sessions sharp.
- Permission modes: plan mode, auto mode and more.
- CLAUDE.md guide: project memory that Claude follows.
- Worktrees: run parallel sessions safely.
- VS Code and JetBrains: Claude Code inside your editor.
Extend and automate
- Skills guide and the Agent Skills standard.
- Best Claude Code skills: picks worth installing.
- Subagents guide and the wider Claude agents overview.
- Hooks guide.
- MCP servers for Claude and the best MCP servers.
- Plugins guide and building a plugin marketplace.
- Headless mode and GitHub Actions.
Once you have the basics, you will start collecting your own skills, hooks and servers. That collection is your Claude kit, and our Claude Code tool listing tracks the official project links.
Who Claude Code is for
- Professional developers who want an agent that can take a ticket and produce a tested diff.
- Teams that want shared conventions (
CLAUDE.md, skills, plugins) and automated PR review. - Solo builders and learners who want to move faster and learn by reading the agent’s changes.
If you don’t write code and don’t want a terminal, the Claude desktop app’s Cowork mode offers agentic help for documents and files instead. Either way, the quickest way to judge Claude Code is to install it and give it a small, real task from your own project.
Frequently asked questions
What is Claude Code in simple terms?
Claude Code is Anthropic's agentic coding tool. You describe a task in plain language and it reads your codebase, edits files, runs commands such as tests and git, and reports back. It runs in the terminal, in VS Code and JetBrains, in the Claude desktop app and in the browser at claude.ai/code.
Is Claude Code free?
The software is free to install, but using it requires a paid Claude plan (Pro, Max, Team or Enterprise), a Claude Console account billed per token, or a supported cloud provider such as Amazon Bedrock. The free Claude.ai plan does not include Claude Code.
Is Claude Code the same as Claude?
No. Claude is the family of AI models and the chat app at claude.ai. Claude Code is a separate product built on those models, designed to work inside a software project: it can read and change files and run commands, which the chat app cannot do on your machine.
Do I need to know how to code to use Claude Code?
It helps. Claude Code writes and runs code for you, but you still decide what to build, review its changes and approve risky actions. Non-developers who want agentic help without a terminal often prefer Cowork in the Claude desktop app.
How do I install Claude Code?
On macOS, Linux or WSL run curl -fsSL https://claude.ai/install.sh | bash, on Windows PowerShell run irm https://claude.ai/install.ps1 | iex, then run claude inside a project folder and sign in. Homebrew and WinGet packages are also available.