Claude Code subscription: plans, prices and how to install it (2026)

Plans, prices and features checked on the official pages on October 17, 2026.

Claude Code is included in every paid Claude plan: Pro costs $17 a month on annual billing ($20 billed monthly) and Max starts at $100 a month. The free Claude plan does not include it, but a pay-as-you-go Claude Console account also works. Install it with one command from a terminal, then type claude in a project folder.

What Claude Code is and who it is for

Claude Code is an agentic coding tool made by Anthropic. According to Anthropic, you hand it a task such as a bug fix, a test or a multi-day migration, and it reads your codebase, edits files, runs commands and can open pull requests. It asks permission before changing files or running commands.

It is aimed at developers, but the docs note that it can help with anything you can do from the command line, including writing docs and searching files. Anthropic owns and develops it. Anthropic’s release notes date the research preview to February 24, 2025 and general availability to May 22, 2025.

How Claude Code works

Anthropic describes an agentic loop with three phases: gather context, take action, and verify results. Claude uses tools throughout: it reads and edits files, searches code, runs shell commands and git, and can search the web and fetch documentation. You can interrupt at any point to steer it.

The docs list five built-in tool groups: file operations, search, execution, web, and code intelligence (which needs a plugin). Claude Code runs locally and talks directly to Anthropic’s model APIs, according to the product page, with no backend server or remote code index.

Before Claude edits a file it saves a snapshot, so pressing Esc twice rewinds to an earlier state. Checkpoints cover file changes only, not actions on remote systems such as databases or deployments.

Key features

• Plan mode: Claude explores and proposes a plan without editing source files. Press Shift+Tab until the status bar shows plan mode, or start with the option –permission-mode plan.
• Permission modes: Auto (a classifier blocks risky actions), Manual, Accept edits and Plan.
• CLAUDE.md and memory: a markdown file in your project that Claude reads at the start of every session. Claude can also save notes through auto memory.
• Skills: packaged, repeatable workflows your team can share, such as a PR review routine. They load on demand.
• Hooks: shell commands that run before or after Claude Code actions, for example auto-formatting after every edit.
• MCP: the Model Context Protocol connects Claude Code to tools like Jira, Slack or Google Drive.
• Subagents and the Agent SDK: parallel agents for parts of a task, and an SDK for building your own agents on the same tools.
• Git and GitHub: Claude Code stages changes, writes commit messages, creates branches and opens pull requests. GitHub Actions and GitLab CI/CD integrations automate review and triage.
• Images: you can paste, drag in or point to an image, and Claude can analyze screenshots, diagrams and mockups.

Anthropic’s pages do not list PDF reading or image generation as Claude Code features. They describe image analysis only, so treat those two as unconfirmed.

Where you can use it

Every surface uses the same engine, so your CLAUDE.md files, settings and MCP servers carry over.

• Terminal (CLI): the full-featured version.
• IDEs: extensions for VS Code (also Cursor) and a JetBrains plugin. The JetBrains plugin needs the CLI installed separately.
• Desktop app: for macOS, Windows (x64 and ARM64) and a Linux beta. It includes Claude Code, so no separate CLI is needed, and it requires a paid subscription.
• Web and mobile: claude.ai/code and the Claude apps for iOS and Android.
• Slack and CI: mention @Claude in Slack with a bug report, or run it in GitHub Actions.

Claude Code subscription plans and pricing

Prices checked on October 17, 2026. Max is billed monthly only, and the Max 5x and 20x prices come from the Claude Code product page.

Limits reset on a rolling five-hour window, and paid plans add weekly limits. Claude on web, desktop and mobile shares one usage pool with Claude Code. Anthropic lists Pro for short sprints in small codebases, so heavy users are pointed toward Max or the API.

Paying per token instead

With a Claude Console account, Claude Code bills API tokens at standard rates. Anthropic’s pricing page lists these per million tokens (input / output):

• Opus 5.5: $4 / $20
• Sonnet 5.5: $2 / $10
• Haiku 5.5: $0.10 / $0.50

For enterprise deployments, Anthropic’s cost documentation gives an average of about $13 per developer per active day and $150 to $250 per developer per month. It says costs stay below $30 per active day for 90% of users. Costs vary with model, codebase size and habits. Subscribers can add usage credits to keep working past their plan limit, using the /usage-credits command.

Anthropic’s product page also lists a fast mode for Opus 5.5, described as 2.5x faster at a higher price of $8 / $40 per million tokens, in research preview.

How to install Claude Code

Anthropic recommends the native installer. Requirements per the docs: macOS 13+, Windows 10 1809+ or Windows Server 2019+, Ubuntu 20.04+, Debian 10+ or Alpine 3.19+, with 4 GB of RAM and an internet connection. Use is limited to Anthropic supported countries; see also why access to Anthropic’s newest models is a live issue.

1. Get a Pro, Max, Team, Enterprise or Console account.
2. Open a terminal. On macOS, Linux or WSL, run the shell installer: curl -fsSL https://claude.ai/install.sh | bash
3. On Windows PowerShell, run: irm https://claude.ai/install.ps1 | iex. Users of the older Command Prompt (CMD) will find a separate batch-file installer in the official setup guide.
4. Open a new terminal window and check claude –version. A version number means it worked.
5. Go to your project folder, start claude, and log in through the browser prompt.

Other routes: Homebrew (brew install –cask claude-code) on macOS, WinGet (winget install Anthropic.ClaudeCode) on Windows, apt, dnf or apk repositories on Linux, or npm (npm install -g @anthropic-ai/claude-code), which needs Node.js 22 or later. If you prefer a graphical app, download the desktop app instead and open its Code tab.

Windows notes

Claude Code runs natively on Windows or inside WSL, and you do not need Administrator rights. Git for Windows is optional: with it, Claude uses Git Bash, and without it, Claude uses PowerShell as its shell tool. Sandboxing is supported on WSL 2 but not on native Windows.

Updates, versions and status

Native installs update themselves in the background. Homebrew, WinGet and Linux package manager installs do not, so upgrade them yourself with brew upgrade claude-code or winget upgrade Anthropic.ClaudeCode. To update right away, run claude update. The docs show a version number such as 2.1.211 as an example output of claude –version, and claude doctor prints installation diagnostics.

You can choose the latest channel (the default) or stable, which is typically about a week behind and skips releases with major regressions. Release history lives in Anthropic’s Claude Code docs and release notes. To check whether Claude Code is down, see status.claude.com, which lists Claude Code as its own component. It showed Operational on October 17, 2026.

Limits and things to know

• The free Claude plan has no Claude Code access.
• Long sessions use more of your limit, because Claude Code sends the full conversation with each request. Anthropic suggests /clear between unrelated tasks and Sonnet instead of Opus for most work.
• A session limit message applies across all models on a subscription, so switching models does not restore access. The exception is a model-specific limit message.
• Permission modes matter: Auto mode starts by default in newer versions, while Manual asks before every edit.
• Agent teams are off by default and use roughly 7x more tokens than standard sessions when teammates run in plan mode, according to Anthropic.

On the 2026 source code leak: according to The Register and The Hacker News, version 2.1.88 of the npm package shipped with a source map on March 31, 2026, exposing about 512,000 lines of the CLI’s TypeScript. Anthropic called it a release packaging error, not a security breach, and said no customer data or credentials were exposed. The model itself was not part of it.

Claude Code next to other tools

Claude Code is not a single editor. It runs in a terminal, in the Cursor and VS Code editors as an extension, and in JetBrains IDEs, so it can sit beside an editor like Cursor rather than replace it. Comparisons with OpenAI’s Codex or Cursor’s own agent depend on those makers’ current plans and limits, which are outside this guide. Check each maker’s pricing page before deciding.

Anthropic’s growth shows how much weight it places on this product; see Anthropic’s enterprise revenue surge and its report on AI agents working far more hours than staff.

Common questions