You can sign in to Claude Code on a remote computer even when that computer has no usable browser. Start authentication in the remote terminal, open its login URL in your own browser, and return to the same terminal to finish.
This is interactive remote login, not an unattended authentication setup. You still need an eligible account and an authorized connection to the computer. A rented Mac does not change account eligibility.
Keep the browser, terminal, and execution computer separate
Before starting, identify three things:
An SSH connection does not make your laptop's Claude login available to the remote process. Likewise, signing in inside a browser does not by itself prove that the remote CLI completed authentication.
If you are still choosing where to run tasks, start with Claude Code on Agent.Space. The steps below assume the CLI is installed and you already have approved SSH access.
Start from the remote terminal
In the SSH terminal, check the machine and directory:
These commands identify the environment and start Claude Code. They do not copy your project from your laptop. Check any workspace-trust prompt before proceeding.
Use the sign-in prompt, or /login if you need to sign in from an existing session. If Claude is already authenticated, inspect /status before changing anything: you may already be using the intended account.
An API credential can select a different access route from the subscription you expected. Check the active route without printing secret values. On an organization-managed machine, ask the administrator before changing credentials or settings.
Complete login in your own browser
The official authentication instructions describe copying the login URL with c when the browser does not open. Open that URL in your local browser and check which account you are authorizing.
If the browser returns a login code rather than completing the callback, paste it into the matching prompt in the remote terminal. Keep the original terminal open while doing this. Use the code only in that login flow; do not send it to a teammate or paste it into an Agent conversation.
Confirm that the terminal reports successful login. Then inspect /status and the project directory before sending a real task. Signing in as the wrong account can change billing and permissions even when the login itself succeeds.
If the browser finishes but the terminal does not
First locate the point of failure:
- The terminal still expects a code: return to the browser's completion screen and follow the matching prompt.
- You restarted login midway: use the new flow rather than mixing an old browser tab with a new terminal prompt.
- The account is already signed in but access later expires: follow the login-expiration guide.
- The notice asks for identity verification, account review, or organization access: address that notice. A different browser or computer is not a substitute for authorization.
Do not copy another person's credential files or upload your whole Claude configuration directory to a repository. For support, share the redacted error and the step that failed, not the login URL, code, or token.
These are documentation-based instructions; we have not completed authentication on your remote account.
Connect the working computer to Agent.Space
If your goal is to keep using an already configured Mac from an Agent.Space Workspace, follow the computer and local-Agent connection guide. Connect the intended folder, confirm that the Mac App detects Claude Code, and select the path that uses that computer's Agent and subscription.
For the first task, ask it to identify the computer and list the project files without editing anything. Verify those details before authorizing changes. Local-Agent use and a cloud Agent accessing a local folder are different options; check the selected route and its billing.
Need to operate a graphical login window on that Mac instead? Review Remote Screen. It provides screen access, not an exemption from the model provider's account requirements.