> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/ai-workbench/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/ai-workbench/_mcp/server.

# Install Claude Code in a Project Container

Follow these instructions to install Claude Code in an AI Workbench project container to persist it across container restarts.

In this quickstart, you will:
\- Install Claude Code in the project container via the `postBuild.bash` script
\- (Optionally) Add Claude Code to the JupyterLab Launcher
\- Start the container and verify Claude Code was installed

## Prerequisites

**To do this quickstart you should already have:**

> 1. The AI Workbench Desktop Client installed locally in full local mode (or an existing remote location)
> 2. An existing project to add Claude Code to
> 3. An Anthropic account with an API key or authentication method ready

## Add Claude Code to the Build Script

**Step One: Open the postBuild.bash file in a file editor.**

1. Select **Project Tab > File Browser > Option Dots > Edit File** for `postBuild.bash`

**Step Two: Add dependency and installation commands to the script.**

Add the following to the script.

```bash
# Install apt packages used by Claude Code features
# bubblewrap: process sandbox (reduced capability in containers)
# socat: proxied network access for domain allowlists
# jq: useful for writing hooks directly in settings.json
sudo apt-get update && sudo apt-get install -y jq bubblewrap socat

# Install Claude Code
curl -fsSL https://claude.ai/install.sh | bash
```

## (Optional) Add Claude Code to the JupyterLab Launcher

## Build and Verify

**Step One: Build the container.**

1. Select **Project Tab > Project Container > Build**
2. Wait for the build to complete

**Step Two: Open a terminal in the container.**

1. Open JupyterLab or attach a terminal to the running container

**Step Three: Verify Claude Code is installed.**

1. Run `claude --version` in the terminal
2. Run `claude` to start a session and authenticate when prompted

**Success: Claude Code is installed in the container image and available on every container start.**

## Next Steps

- [Configure Claude Code Sandboxing in a Project Container](/ai-workbench/quickstart/quickstart-claude-sandbox) - Create a sandbox configuration repository for Claude Code
- [Sandboxing Agents with AI Workbench](/ai-workbench/concepts/sandboxing-concept) - Background on how agent sandboxing works in AI Workbench