What Claude Code does.
Claude Code is an Anthropic tool that understands an entire project and acts on it: it reads the files, plans an approach, writes and edits code across several files, runs commands and checks that the result works. It also works directly with Git: it stages changes, writes commit messages, creates branches and opens pull requests.
The official documentation lists the tasks it handles day to day: writing tests for untested code, fixing errors across a whole project, resolving merge conflicts, updating dependencies, writing release notes. You describe what you want in plain language; it takes care of it.
That is the key difference with a chat: a chat advises, Claude Code does. For the broader distinction, see AI Agent vs Chatbot.
Install Claude Code in one command.
The official documentation gives one command per system:
# macOS, Linux, WSL
curl -fsSL https://claude.ai/install.sh | bash
# Windows PowerShell
irm https://claude.ai/install.ps1 | iex
Other routes exist: Homebrew on Mac (brew install --cask claude-code), WinGet on Windows (winget install Anthropic.ClaudeCode), or Linux package managers. The native install updates itself in the background, which Homebrew and WinGet installs do not.
Then check that everything works and start in a project:
claude --version
cd your-project
claude
On first launch, Claude Code asks you to sign in. On Windows, the documentation recommends installing Git for Windows so it can use its Bash tool.
Terminal, desktop, IDE, browser: where to use it.
| Surface | For whom | Strength |
|---|---|---|
| Terminal (CLI) | Technical profiles | The most complete version |
| Desktop app | Everyone | Code tab, parallel sessions, visual change review, scheduled tasks |
| VS Code and JetBrains | Developers | Diffs shown in the editor, conversation history |
| Browser (claude.ai/code) | Remote work | Cloud sessions, no install, on GitHub repositories |
The desktop app includes Claude Code: no need to install the command line separately. It is often the easiest entry point for a team that does not live in a terminal.
Which plan you need for Claude Code.
Claude Code requires a paid account. The plans that include it, as shown on claude.com on September 27, 2026 (prices for France, excluding tax):
| Plan | Price | For whom |
|---|---|---|
| Pro | €15/month billed annually (€180 upfront), €18 monthly | Regular individual use |
| Max | From €90/month | Heavy use, 5 or 20 times Pro's usage |
| Team | €18/seat/month annually for a Standard seat, €90 for a Premium seat | Teams of 2 to 150 |
| Enterprise | $20/seat/month billed annually, plus usage at API rates | Large organizations |
You can also use Claude Code with a Claude Console account and API credits, paying per use. Claude's free plan does not include Claude Code. Each plan's usage limits are detailed in Claude Pricing 2026.
Skills, MCP, subagents: extending it.
Claude Code becomes truly useful once adapted to your context. The documentation describes several mechanisms:
- CLAUDE.md: an instruction file read every session. Project rules, conventions and known pitfalls go there.
- Skills: reusable procedures in a
SKILL.mdfile, which Claude uses when relevant or you call with/skill-name. Everything is explained in Claude Skills. - Subagents: specialized agents Claude delegates part of the work to, in their own isolated context.
- Hooks: actions triggered automatically at given points in Claude Code's lifecycle.
- MCP: the connection to your tools and data through MCP servers. See MCP (Model Context Protocol).
- Plugins: packages bundling skills, subagents, hooks and MCP servers, to share them.
Beyond code: what businesses do with it.
Claude Code is called "Code", but its real talent is carrying out multi-step tasks on files and tools. That makes it useful far beyond developers.
At AUTOMATE ALL, we use it every day. Preparing our prospecting messages runs as a skill: from a name, Claude reads the prospect's profile and recent activity, then drafts a tailored approach we review before sending. The same principle serves to produce documents, analyze exports, build internal tools or prepare automations.
Two precautions for business use:
- Stay in control of what goes out. Anything that sends, publishes or deletes goes through human approval.
- Scope the access. Claude Code acts with the rights you give it: start small, on a specific folder and specific tools.
To go further on AI agents in business, see our services.
Claude Code: what people ask us.
Is Claude Code free?
No. Claude's free plan does not include it. You need a Pro subscription (€15/month billed annually, €18 monthly, as shown for France), Max, Team or Enterprise, or a Claude Console account with API credits. Plan details are in Claude Pricing 2026.
How do I install Claude Code?
On macOS, Linux or WSL: curl -fsSL https://claude.ai/install.sh | bash. On Windows PowerShell: irm https://claude.ai/install.ps1 | iex. Check with claude --version, then run claude in your project folder and sign in on first launch.
Do you need to be a developer to use Claude Code?
It helps, because it works in folders and files, often from a terminal. But the desktop app and the browser version lower the barrier a lot. We also use it for tasks that are not code: preparing documents, processing files, automating repetitive tasks.
What is the difference between Claude Code and the Claude chat?
The chat answers in a conversation. Claude Code acts in a project: it reads and edits your files, runs commands and checks the result. It is the difference between an assistant that advises and an agent that does.
Can Claude Code connect to our tools?
Yes, with MCP (Model Context Protocol): you add MCP servers so it can reach your tools and data. You can also extend it with skills, subagents, hooks and plugins.
Where does Claude Code run?
In the terminal, in VS Code and JetBrains IDEs, in the Claude desktop app (Code tab), and in the browser at claude.ai/code, where sessions run in the cloud with no local install.