Run Claude Code in a Daytona Sandbox via CLI

This guide walks you through running Claude Code inside a Daytona sandbox using the Daytona CLI.

Prerequisites

• Daytona account and API key (Get it from Daytona Dashboard)
• Local terminal (macOS, Linux, or Windows)

Install the Daytona CLI

Mac/Linux

brew install daytonaio/cli/daytona

Windows

powershell -Command "irm https://get.daytona.io/windows | iex"

Note

Already have the CLI? Check your version with daytona --version. If it’s below 0.135.0, upgrade to the latest version.

Authenticate with Daytona

Log in to your Daytona account using your API key:

daytona login --api-key=YOUR_API_KEY

Replace YOUR_API_KEY with your actual Daytona API key.

Create a Sandbox

Create a new sandbox for running Claude Code:

daytona sandbox create --name claude-sandbox

This creates a sandbox named claude-sandbox, visible in your Dashboard. The default Daytona snapshot includes Claude Code, so the command above is all you need.

Tip

Need more power? Pass --snapshot daytona-large or --snapshot daytona-medium flag to increase your sandbox resources. See default snapshots for resource details.

Connect to the Sandbox

SSH into your sandbox:

daytona ssh claude-sandbox

You now have an interactive terminal session inside the sandbox.

Run Claude Code

Inside the SSH session, start Claude Code:

claude

On first run, Claude Code will prompt you to authenticate:

1. Copy the authentication URL displayed in the terminal
2. Open the URL in your local browser
3. Complete the authentication flow
4. Copy the code provided by the browser
5. Paste the code back into the terminal

Once authenticated, you’re all set. Claude Code runs inside the sandbox while you control it from your terminal.