- Configure devcontainer with Node.js 20, Claude Code CLI, and dev tools - Add firewall script to restrict network access to whitelisted domains - Create run-container.sh helper for interactive and non-interactive usage - Support interactive mode (no args) for authentication/credential storage - Support non-interactive mode with prompt argument or stdin 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
134 lines
4.3 KiB
Markdown
134 lines
4.3 KiB
Markdown
# CLAUDE.md
|
|
|
|
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
|
|
|
|
## Development Container Setup
|
|
|
|
This repository is configured with a secure development container based on Anthropic's reference devcontainer setup. The container provides:
|
|
|
|
- **Isolated environment**: Containerized Claude Code CLI with firewall restrictions
|
|
- **Security features**: Network access limited to whitelisted domains (GitHub, npm, Anthropic APIs, etc.)
|
|
- **Pre-configured tools**: Node.js 20, Claude Code CLI, git-delta, zsh with powerline10k, fzf, and more
|
|
- **Persistent storage**: Volumes for bash history and Claude configuration
|
|
|
|
### Opening in DevContainer
|
|
|
|
**Option 1: Using run-container.sh script (easiest for non-interactive usage)**
|
|
|
|
The repository includes a helper script for running Claude Code non-interactively:
|
|
|
|
```bash
|
|
# Make the script executable (first time only)
|
|
chmod +x run-container.sh
|
|
|
|
# Run with a prompt
|
|
./run-container.sh "explain the fibonacci sequence"
|
|
|
|
# Run with stdin
|
|
echo "create a hello world function in Python" | ./run-container.sh
|
|
|
|
# Analyze a file
|
|
./run-container.sh "explain this code" < myfile.js
|
|
```
|
|
|
|
The script automatically:
|
|
- Builds the Docker image if needed
|
|
- Creates persistent volumes for history and config
|
|
- Initializes the firewall
|
|
- Runs claude with streaming output and --dangerously-skip-permissions
|
|
- Cleans up the container after execution
|
|
|
|
**Option 2: Using devcontainer CLI (recommended for interactive development)**
|
|
|
|
Install the devcontainer CLI:
|
|
```bash
|
|
npm install -g @devcontainers/cli
|
|
```
|
|
|
|
Build and run the container:
|
|
```bash
|
|
# Build the container
|
|
devcontainer build --workspace-folder .
|
|
|
|
# Run the container and execute a command
|
|
devcontainer exec --workspace-folder . claude -p "your prompt" --dangerously-skip-permissions
|
|
|
|
# Or open an interactive shell
|
|
devcontainer exec --workspace-folder . zsh
|
|
```
|
|
|
|
**Option 3: Using Docker directly**
|
|
|
|
Build and run manually:
|
|
```bash
|
|
# Build the image
|
|
docker build -t claude-dev-container .devcontainer
|
|
|
|
# Create volumes for persistence
|
|
docker volume create claude-code-bashhistory
|
|
docker volume create claude-code-config
|
|
|
|
# Run interactively
|
|
docker run -it --rm \
|
|
--cap-add=NET_ADMIN \
|
|
--cap-add=NET_RAW \
|
|
-v "$(pwd):/workspace" \
|
|
-v claude-code-bashhistory:/commandhistory \
|
|
-v claude-code-config:/home/node/.claude \
|
|
-e NODE_OPTIONS="--max-old-space-size=4096" \
|
|
-e CLAUDE_CONFIG_DIR="/home/node/.claude" \
|
|
-w /workspace \
|
|
--user node \
|
|
claude-dev-container zsh
|
|
|
|
# Inside the container, initialize the firewall:
|
|
sudo /usr/local/bin/init-firewall.sh
|
|
|
|
# Then use Claude Code:
|
|
claude -p "your prompt" --dangerously-skip-permissions
|
|
```
|
|
|
|
**Option 4: VS Code**
|
|
1. Install the "Dev Containers" extension
|
|
2. Open this repository in VS Code
|
|
3. When prompted, click "Reopen in Container" (or use Command Palette: "Dev Containers: Reopen in Container")
|
|
4. Wait for the container to build and the firewall to initialize
|
|
|
|
**First-time setup:**
|
|
- You'll need to authenticate Claude Code on first use
|
|
- Run `claude` in the container terminal and follow the authentication prompts
|
|
|
|
### Running Claude Code Non-Interactively
|
|
|
|
The devcontainer's firewall allows running Claude Code with `--dangerously-skip-permissions` for non-interactive operation:
|
|
|
|
```bash
|
|
# Stream output in non-interactive mode
|
|
claude -p "your prompt here" --dangerously-skip-permissions
|
|
|
|
# Example: Analyze a file
|
|
claude -p "explain this code" --dangerously-skip-permissions < myfile.js
|
|
|
|
# Example: Generate code with streaming output
|
|
echo "create a fibonacci function" | claude -p --dangerously-skip-permissions
|
|
```
|
|
|
|
**Security notes:**
|
|
- The `--dangerously-skip-permissions` flag bypasses permission prompts
|
|
- This is safe within the devcontainer due to firewall restrictions
|
|
- Network access is limited to: GitHub, npm registry, Anthropic APIs, VS Code services
|
|
- All other outbound connections are blocked
|
|
|
|
### Container Configuration Files
|
|
|
|
- `.devcontainer/devcontainer.json` - Container and VS Code configuration
|
|
- `.devcontainer/Dockerfile` - Container image definition
|
|
- `.devcontainer/init-firewall.sh` - Network security rules (runs on container start)
|
|
|
|
### Environment Setup (Outside Container)
|
|
|
|
This repository uses [mise](https://mise.jdx.dev/) for tool version management.
|
|
|
|
- Node.js version: 24 (configured in `mise.toml`)
|
|
- Install tools: `mise install`
|