How to Configure Claude Code settings.json in AI Skill Notebook

When you use the Claude Code CLI, it will often ask for your approval before it makes network requests or edits files. If you want Claude Code to run fully automatically—or you want to grant it access to additional tools—you’ll need to configure the appropriate rules in Claude Code’s settings.

The settings.json file

If you’re using AI Skill Notebook, you can configure Claude Code by editing:

<your-workspace>/.claude/settings.json

In AI Skill Notebook, you can open and edit this file via:

Settings (top-right) → Workspace Settings → Claude Code Configuration → Open

Note
Google Gemini and Claude Code share the same settings.json.
In practice, your real settings.json may contain configuration for both Gemini and Claude Code.

Example settings.json

{
  "permissions": {
    "defaultMode": "acceptEdits"
  },
  "sandbox": {
    "enabled": false,
    "autoAllowBashIfSandboxed": true
  }
}

permissions.defaultMode

This controls whether Claude Code should ask for confirmation before taking actions.

AI Skill Notebook sets the default to acceptEdits, but you can change it to one of the following values:

• default — Ask before using each tool

• acceptEdits — Automatically allow generally safe actions (especially file edits)

• plan — Read-only / planning mode (no modifications)

• dontAsk — Do not authorize actions and do not ask for permission

• bypassPermissions — Allow everything (use with caution)

sandbox

This controls whether sandbox mode is enabled.

In sandbox mode, Claude Code won’t directly operate on your real files—it will simulate file operations instead.

• enabled: false — Claude Code works on real files

• enabled: true — Claude Code operates in a sandbox (safer, but less direct)

autoAllowBashIfSandboxed determines whether shell commands are automatically permitted when sandbox mode is enabled.

Using third‑party MCP tools (Model Context Protocol)

Third‑party software that supports MCP (Model Context Protocol) can be controlled through natural language.

Take Playwright as an example:

• Developers can use Playwright via code to automate a browser.

• Playwright MCP allows Claude Code to control the browser through natural language instructions, interpreting your intent and executing browser actions.

Enable Playwright MCP in Claude Code

To use Playwright MCP in Claude Code, create a file named .mcp.json in your project root and add:

{
  "playwright": {
    "command": "npx",
    "args": ["@playwright/mcp@latest"]
  }
}

More settings

For a full list of options, see the official documentation:

• Claude Code settings: https://code.claude.com/docs/en/settings