A containerized environment for testing and experimenting with Claude Code features in isolation, without affecting your main configuration.
This sandbox provides a Docker-based environment where you can safely test Claude Code's features. Your authentication and configuration persist between sessions, and since the project directory is mounted, you can modify .claude/ settings without rebuilding the container.
- Custom Commands: Create slash commands in
.claude/commands/to automate repetitive tasks - Skills: Build reusable AI skills in
.claude/skills/for specialized workflows - Hooks: Configure hooks in
.claude/claude.jsonfor automation - Subagents: Experiment with delegating tasks to specialized Claude agents
- MCP Servers: Test Model Context Protocol integrations
-
Clone this repository
git clone https://github.com/fz42net/claude-code-sandbox cd claude-code-sandbox -
Build and start the container
docker compose up -d
-
Enter the container
docker compose exec claude-code-sandbox zsh -
Login to Claude Code (first time only)
claude # Follow the authentication prompts -
Start experimenting!
- Modify
.claude/commands/,.claude/skills/, or.claude/claude.json - Changes take effect immediately - no rebuild needed
- Exit the container with
exit, your auth persists for next time
- Modify
.claude/
├── commands/ # Custom slash commands
├── skills/ # Reusable AI skills
└── claude.json # Configuration and hooks (create as needed)
For comprehensive documentation on Claude Code features: https://docs.claude.com/en/docs/claude-code