> ## Documentation Index
> Fetch the complete documentation index at: https://mintlify.com/killlowkey/claude-code/llms.txt
> Use this file to discover all available pages before exploring further.

# Settings

> Configure Claude Code behavior through the /config command and configuration files.

Claude Code stores configuration in `~/.claude/` and supports both interactive configuration via `/config` and direct file editing.

## Interactive configuration

The `/config` command opens an interactive settings panel where you can browse and change all settings:

```
> /config
```

The panel lets you navigate settings with arrow keys and toggle or edit values interactively. It is also aliased as `/settings`.

## Configuration options

### Model selection

Switch models with `/model`:

```
> /model
```

This opens a model picker showing available Claude models. Available models depend on your Anthropic subscription.

You can also specify a model at startup:

```bash theme={null}
claude --model claude-sonnet-4-6    # full model ID
claude --model sonnet               # alias for the latest Sonnet
claude --model opus                 # alias for the latest Opus
```

### Permission mode

Controls how Claude handles tool permission prompts:

| Mode                | Behavior                                                |
| ------------------- | ------------------------------------------------------- |
| `default`           | Prompts for approval on each new tool/pattern           |
| `plan`              | Plans actions but requires explicit approval to execute |
| `auto`              | Auto-approves based on learned patterns                 |
| `bypassPermissions` | Skips all permission checks (use with caution)          |

Set via the `--permission-mode` CLI flag at startup, or changed interactively from the `/config` panel.

### Theme

Change the terminal color theme:

```
> /theme
```

Themes affect syntax highlighting and UI element colors throughout the REPL.

### Vim mode

Enable vim keybindings for the input prompt:

```
> /vim
```

Toggle off by running `/vim` again.

### Effort level

Adjust how much reasoning Claude applies:

```
> /effort
```

Higher effort enables extended thinking for complex architectural decisions. Lower effort gives faster responses for simple tasks.

### Output style

Change how Claude formats responses:

```
> /output-style
```

## CLI flags

Pass flags when launching Claude Code:

```bash theme={null}
# Skip permission prompts (use carefully)
claude --dangerously-skip-permissions

# Start with a specific model
claude --model claude-sonnet-4-6

# Non-interactive: print response and exit
claude -p "Summarize the main.ts file"

# Continue the most recent conversation
claude --continue
```

## Environment variables

| Variable                     | Description                                     |
| ---------------------------- | ----------------------------------------------- |
| `ANTHROPIC_API_KEY`          | API key (overrides keychain)                    |
| `ANTHROPIC_MODEL`            | Default model to use (e.g. `claude-sonnet-4-6`) |
| `HTTP_PROXY` / `HTTPS_PROXY` | Proxy server for all requests                   |
| `NO_PROXY`                   | Hosts to bypass proxy                           |

<Note>
  On macOS, Claude Code uses the system Keychain to store credentials securely. The `ANTHROPIC_API_KEY` environment variable overrides the Keychain value if set.
</Note>

## Project-level configuration

Place a `CLAUDE.md` file in your project root to give Claude persistent context about your project:

```markdown theme={null}
# My Project

## Tech stack
- Node.js 20, TypeScript, PostgreSQL
- Testing: Vitest

## Conventions
- Use named exports, not default exports
- All async functions must handle errors explicitly
- Database queries go through the repository layer only
```

Claude reads `CLAUDE.md` automatically at the start of every session in that directory.