Skip to main content
AI models & keys (Settings > AI models & keys) lets your own coding tools and scripts use your company’s models, as you, within your budgets. The key covers model calls only. File edits, shell commands and local skills stay on your computer and are not recorded.

​
Connect a coding tool

1

Pick your client

Under Connect a client, pick Claude Code, Codex CLI, OpenCode or Your own code.
2

Copy and run the setup

  • Claude Code, Codex CLI or OpenCode: pick macOS, Linux or Windows, select Copy (it creates a key for that client and computer), and run it in your terminal (PowerShell on Windows). The coding tool stays connected in every new terminal.
  • Your own code: select Copy and paste it into your project’s coding agent. It wires your OpenAI or Anthropic SDK calls to the gateway, reading the key from XPANDER_GATEWAY_KEY. The prompt never contains your key.
Connect a client with Claude Code, Codex CLI, OpenCode and Your own code, Claude Code selected, and Linux picked in the line Copy creates your API key automatically

Connect a client, with Claude Code and Linux picked. The setup command and Copy are below it.

Set up manually shows the files instead, with your address, a new key and a Model filled in.

​
Create a key yourself

Under Gateway keys, choose Create key, give it a name such as My desktop, and pick when it expires (7 days to Never). Copy it: it is Shown only once. Use one key per computer and coding tool. Revoke stops a key at once.

​
Available models

Available models lists what you may call. Copy a model ID and use it as model in your coding tool. Shortcuts your admin set up work the same way. A missing model is turned off for you: ask an admin. If you use your own Claude plan here, set When my Claude plan hits its limit to Continue on company models or Pause, unless your admin set calls to pause. Your spend is in Usage (Settings > Usage).

​
Next steps

AI Gateway

What your admins control.

Gateway errors

What a refused call means.