Multi-Client Wireup - ChatGPT, Claude Code, Grok, Copilot

One server, every client. The same node you stood up works with all of the major AI clients, and you connect each one only once. Because they all speak the same protocol, there is nothing client-specific on the server; the only difference is the shape of the values each one asks for.

Claude

Claude runs the sign-in itself, so it needs the least: the server address plus an id and secret. Paste those into its connector and approve when it asks.

ChatGPT and Grok

These want the addresses spelled out: the server, the sign-in address, and the token address, plus the id and secret and a scope. It is a few more fields, but it is the same node behind them.

Command-line tools (Claude Code, Codex, Copilot)

These use a token directly, with no sign-in dance. You point them at the server address and hand them the token. The exact per-client fields are in the connect doc.

The unlock

Here is the part that surprises people. Once it is connected, you are not tied to a terminal. From ChatGPT or Claude in a browser, you can manage and deploy your cloud, which means you can do it from your phone. Standing up infrastructure from your couch, on mobile, is the kind of thing that used to take a laptop and a VPN. Now it is a chat.