Configuration Commands

Open and navigate the settings interface with /config

Switch between AI models with /model

Understand when to use different models

Manage permissions with /permissions

Toggle features like fast mode and sandboxing

Claude Code is highly configurable. You can change which AI model powers your session, adjust what Claude is allowed to do, and toggle features that affect speed and safety. These configurations happen through slash commands.

This lesson covers the main configuration commands. You will learn to open the settings interface, switch models for different tasks, manage permissions, and enable features that match your workflow. By the end, you will be able to customize Claude Code to work the way you prefer.

The /config command opens the main settings interface:

This brings up a menu where you can adjust various options. The settings are organized into categories:

General settings include theme, model selection, output style, and editor mode (normal or Vim keybindings).

Privacy settings control data handling. These are only available on Pro and Max plans.

Thinking settings control extended thinking behavior. You can enable or disable visible thinking summaries.

You can also access settings with the alias /settings.

Navigate the settings interface with arrow keys and Enter. Changes take effect immediately. Some settings persist across sessions (stored in your home directory), while others are session-specific.

The settings file is stored at ~/.claude/settings.json. You can edit this file directly if you prefer, but /config provides a friendlier interface.

Claude Code can use different AI models. Each model has different characteristics:

Claude Opus 4.6 is the most powerful model. It has the deepest reasoning capabilities and handles complex tasks best. It is also slower and uses more tokens.

Claude Sonnet 4.6 balances capability and speed. It is fast enough for interactive use while being powerful enough for most coding tasks. This is the default for most users.

Claude Haiku 4.5 is the fastest and cheapest model. It works well for simple tasks where you do not need deep reasoning.

To switch models:

This opens a picker. Use arrow keys to navigate and Enter to select.

You can also adjust the effort level for models that support it. Effort controls how much thinking the model does before responding:

Low: Quick responses, minimal thinking

Medium: Balanced (default)

High: More thorough reasoning

Max: Maximum reasoning depth (Opus 4.6 only)

In the /model picker, use left and right arrows to adjust effort level for the selected model.

When to switch models:

Use Opus for complex architectural decisions, difficult bugs, or tasks requiring deep reasoning

Use Sonnet for everyday coding work (the default)

Use Haiku for simple questions, quick lookups, or when you want speed over depth

You can also set effort directly:

This changes the effort level without switching models.

The /permissions command controls what Claude is allowed to do:

This opens an interactive dialog where you can:

View current permission rules

Add allow rules for specific tools

Add deny rules to block certain actions

Manage working directories

Review recent auto mode denials

Permission scopes:

Rules can be set at different levels:

Session: Applies only to the current session

Project: Applies to this project directory

User: Applies to all your projects

Common permission patterns:

Allow a specific command without asking:

Allow commands matching a pattern:

Deny dangerous commands:

We cover permissions in more detail in a later lesson. For now, know that /permissions is where you control what Claude can do automatically versus what requires your approval.

Fast mode trades some capability for speed:

When fast mode is on, Claude Code uses optimizations that make responses quicker. This is useful when you are doing rapid iteration and do not need maximum quality on every response.

Fast mode is a toggle. Run /fast without arguments to toggle between on and off, or specify on or off explicitly.

Sandbox mode provides extra isolation for Claude's actions:

When sandboxing is enabled, Claude Code runs in a more restricted environment. This is useful when:

Working with untrusted code

Experimenting with risky operations

Wanting extra protection against accidental damage

Sandbox availability depends on your platform. Not all operating systems support full sandboxing. Run /sandbox to see what is available on your system.

Customize how Claude Code looks:

Change theme:

This opens a theme picker with options including:

Light and dark variants

Colorblind-accessible (daltonized) themes

ANSI themes that use your terminal's color palette

Change prompt bar color:

Available colors: red, blue, green, yellow, purple, orange, pink, cyan. Use /color default to reset.

Configure status line:

This customizes what information appears at the bottom of your screen. You can describe what you want to see, or let Claude Code auto-configure based on your shell.

If the default keyboard shortcuts do not work for you:

This opens your keybindings configuration file where you can remap shortcuts. The file format is JSON with key combinations mapped to actions.

Most users find the defaults work well, but if you have muscle memory from other tools or specific accessibility needs, keybindings can be customized.

Some settings are saved and apply to future sessions:

Theme preference

Model default

Global permission rules

Status line configuration

Keybindings

Other settings are session-specific:

Prompt bar color (set with /color)

Fast mode toggle

Effort level (unless you set it in config)

Session-specific settings reset when you start a new session. Persistent settings stay until you change them.

/config opens the main settings interface

/model lets you switch between Opus, Sonnet, and Haiku models

Use higher-capability models for complex tasks, faster models for simple work

/permissions controls what Claude can do with or without asking

/fast toggles speed optimizations

/theme and /color customize the visual appearance

Settings persist in ~/.claude/settings.json

Explore configuration with these steps:

Start Claude Code and run /config to see the settings interface.

Run /model and browse the available models. Note the effort level indicator.

Run /permissions and look at any existing rules.

Run /theme and try a different color scheme.

Run /color green to change your prompt bar color.

Run /fast on and ask Claude a simple question. Notice the response speed.

Run /fast off and ask the same question. Compare.

Run /effort high and ask a complex question about your code.

This exercise shows how configuration affects your Claude Code experience.

https://code.claude.com/docs/en/settings - Settings interface, settings.json, persistence

https://code.claude.com/docs/en/model-config - Model selection, effort levels, adaptive reasoning

https://code.claude.com/docs/en/permission-modes - Permission modes and configuration