Installation & First Steps
iAs of: September 2026
Model names, prices, and subscription limits reflect the state as of September 2026. Anthropic adjusts these regularly — check claude.ai/pricing before signing up.
Installation
Claude Code can be installed with a single command:
curl -fsSL https://claude.ai/install.sh | bash
iPrerequisites
You need an Anthropic account with a Pro ($20/month), Max, Teams, or Enterprise plan. Claude Code is not available on the Free plan. For API usage, an API key is required; for Pro/Max plans, browser login is sufficient.
After installation, simply start Claude Code with:
claude
This opens interactive mode -- you can start right away and chat with Claude directly in your terminal.
Models
Claude Code uses the latest Claude models. Depending on the task, you can choose the appropriate model:
| Model | Identifier | Strength | Context |
|---|---|---|---|
| Fable 5 | claude-fable-5 | New top class (Mythos) for the hardest tasks | 1M Tokens |
| Opus 5 | claude-opus-5 | Powerful all-round model, complex tasks | 1M Tokens |
| Sonnet 5 | claude-sonnet-5 | Balanced: fast and capable | 1M Tokens |
| Haiku 4.5 | claude-haiku-4-5 | Fast and affordable, ideal for Subagents | 200K Tokens |
*Which Model When?
For most tasks, Sonnet is sufficient. Opus is worth it for complex architecture, large refactorings, or difficult bugs. Haiku is primarily used internally as a subagent for fast, simple subtasks.
Pricing
Claude Code offers various access models:
- Pro ($20/month) -- Shared with Claude Web, Desktop, and Mobile. Good for getting started, but the quota can run tight with intensive use
- Max 5x ($100/month) -- 5x more capacity than Pro. Recommended for regular use
- Max 20x ($200/month) -- 20x more capacity. For power users and professional development
- API -- Pay-as-you-go by tokens. Full cost control, ideal for CI/CD and automation
iCurrent Pricing
Prices and quotas may change. Find the latest details at claude.ai/pricing.
Permission Modes
Claude Code can perform various actions on your system -- read files, write files, execute commands. With Permission Modes, you control how much autonomy you allow:
Default Mode
claude
The standard. Claude Code asks before critical actions -- for example, before overwriting files or executing shell commands. Ideal for getting started.
Accept Edits Mode
claude --acceptEdits
Allows file changes automatically but still asks for shell commands. Good if you trust Claude Code with writing but want to control commands.
Auto Mode
claude --auto
A classifier decides which actions are safe and which need confirmation. Maximum productivity with acceptable risk.
!Security Note
Start in Default Mode and work your way up. Auto Mode is powerful, but you should first develop a feel for how Claude Code works before giving it full autonomy.
iAuto Mode Requirements
The --auto flag requires a Teams, Enterprise, or API plan and works with Sonnet 5 or an Opus model (4.7 or newer).
First Commands
Interactive Mode
claude
The main mode. You chat with Claude Code in an interactive session. Claude sees your project directory and can read files, write files, and execute commands.
Non-interactive Mode
claude -p "Explain the architecture of this project"
Perfect for scripting and CI/CD pipelines. Claude executes the prompt and outputs the result -- without an interactive session.
Resume Last Conversation
claude --continue
Continues the last session. The entire previous context is preserved -- ideal if you interrupted a task.
Important Slash Commands
Within an interactive session, these commands are available:
| Command | Function |
|---|---|
/help | Shows all available commands |
/clear | Clears the current context |
/compact | Compresses the context (saves tokens) |
/init | Creates a CLAUDE.md for your project |
*Pro Tip: /init First
Start every new project with /init. Claude Code analyzes your project and creates a CLAUDE.md file with build commands, architecture overview, and project conventions. This drastically improves all subsequent interactions.