Connecting your coding agent (Claude)

The portal screens are in development. Their words and layout may still change; the commands behind the buttons are already built.

Every workspace comes with the coding agent Claude Code installed. It works under each researcher's own Claude account, so you connect your account once. After that, every workspace uses the connection, and you never sign in workspace by workspace.

What you need

• A Claude subscription. You get the token on your own computer with claude setup-token, which authenticates with your Claude subscription. According to Anthropic's documentation it works with Pro, Max, Team and Enterprise plans, and not with a free claude.ai account.
• Claude Code on your own computer. That command needs Claude Code installed on your computer, not in a workspace.
• A browser. The command asks you to approve in a browser with your Claude account.

1. Install Claude Code on your computer

Skip this if it is already installed: run claude --version in a terminal, and a version number means it is there.

These are the install commands from Anthropic's documentation:

After installing, open a new terminal window so that the claude command is found.

2. Get a token

Run this in a terminal on your own computer:

A browser opens: sign in with your Claude account and approve. The token then appears in the terminal; it starts with sk-ant-.

The command does not save the token anywhere, it only prints it. Treat it like a password: do not leave it in a file or a chat, and paste it in the next step straight away. According to Anthropic's documentation, the token lasts one year.

3. Connect it in the portal

In the menu on the left, open [My account] and go to the Claude section. Paste the token into “Claude token (sk-ant-...)” and press [Connect Claude].

If some of your workspaces are running, you see something like “Connected Claude. Updated running workspaces: ws-0012”. In running workspaces, a Claude panel or claude session opened from now on uses the new token at once, without a restart; one that was already open keeps the old token until you close it and open it again. Stopped workspaces pick it up at their next start.

stdin: (hidden)Connected Claude.Updated running workspaces: ws-0012

If what you pasted is not a token, you see “This is not a Claude token.” with the reason “Paste the token that claude setup-token printed (sk-ant-...).”

Your privacy settings

The Claude section says: “This connects your personal Claude account. Data terms, such as whether conversations may train models, follow your own Claude account settings, and the institution cannot set them. Check your Claude privacy settings before connecting.”

What the agent reads is sent to the model provider (Anthropic) under your Claude account. When the agent reads workspace data, that content is sent too.

Inside a workspace

• The Claude Code panel in the IDE and the claude command in a terminal are already signed in with your token. No sign-in prompt appears.
• The token sits in a read-only file outside the work folder and is handed to Claude Code when it starts (so the commands Claude runs have it too). It is not in the work folder, not in the IDE's or the terminal's environment, and not in the environment saved with a release.
• The token file is readable as the workspace user. The agent's file-reading tool is denied it, but commands the agent runs are not.

Changing the token

Get a new token and connect it the same way. Running workspaces get it at once, but a Claude panel or claude session that was already open keeps the old token until you close it and open it again; do that when you replace an expired token. Today the portal has no button to remove the connection, only to replace the token.

What these docs could not confirm

• Whether the portal warns before the token expires. The screens we checked show only when it was connected.
• How to invalidate a token on Anthropic's side once it has been issued.