Skip to content
dx docs

Use your Claude Pro or Max plan for cloud agents

dx runs Claude Code in cloud sandboxes on the Claude Pro or Max plan you already pay for. You connect the plan once with a token from claude setup-token, then pick Claude Code when you start a Thread. This page covers what the plan pays for, how the token works, and how dx guards it. For a walkthrough of running Claude Code, see Run Claude Code in the cloud.

What the plan covers

Section titled “What the plan covers”

Your plan runs the Claude Code mode in dx, on these models:

Model Note
Claude Opus 5.5 The default
Claude Sonnet 5.5
Claude Haiku 5.5
Claude Fable 5.1
Claude Opus 5
Claude Sonnet 5

You also pick an effort: Low, Medium, High (the default), Extra High, or Max.

The plan does not cover dx’s own modes (Low, Medium, High, Ultra) or raw models. Those run on the other connections in Personal Settings → Model Routing. Claude Code models run only through a Claude Pro/Max connection, so you cannot route them through an API key or a Custom URL.

Get a token with claude setup-token

Section titled “Get a token with claude setup-token”

claude setup-token is a Claude Code command that signs in to your Claude account and prints a long-lived token for that plan.

  1. Install Claude Code on your own computer.

  2. Run:

    Terminal window
    claude setup-token

  3. A browser window opens. Sign in with your Claude Pro or Max account and approve access.

  4. Copy the token the command prints. It starts with sk-ant-oat and is valid for one year.

Connect it to dx

Section titled “Connect it to dx”
  1. In dx, open Personal Settings → Model Routing and choose Add → Claude Pro/Max.
  2. Paste the token and choose Connect Subscription.

dx checks the token with Anthropic before it saves it. If Anthropic rejects it, the dialog says so; run claude setup-token again and paste the new token. Each person connects one Claude plan. To swap the token, choose Replace Token on the connection.

How dx protects the token

Section titled “How dx protects the token”
  • dx stores the token encrypted. Settings shows only that a token is stored, never the token itself.
  • In the sandbox, the token reaches the Claude Code process on a file descriptor for one run, not as a variable in Claude Code’s environment.
  • Before Claude Code starts, dx unsets ANTHROPIC_* and CLAUDE_CODE_USE_* variables and extra CA certificates. Command-line settings, which outrank a repository’s .claude settings, pin Anthropic’s API URL and blank the proxy variables. A repository’s .claude settings or a workspace variable cannot point Claude Code at another API URL or proxy.
  • When Claude Code prints the value of a secret environment variable that dx hides, dx redacts it from the transcript. Settings warns when a secret is too short to hide.

Commands that a repository’s setup or the agent runs in the Orb run as you. Connect your plan only for repositories you trust.

Limits

Section titled “Limits”

Claude Code in dx counts against your plan’s usage limits, the same as Claude Code on your computer. The connection in Model Routing shows your 5-hour, weekly, and per-model limits. By default Claude Code stops at a limit; with extra usage turned on in your Claude account, Allow Extra Usage keeps it working on paid extra usage.

Source

Section titled “Source”