Configure model providers
Connect your Claude and Codex logins, set defaults, and manage org policy.
Kalauz doesn't ask for a second login or its own API key — it drives the claude and codex CLIs already on your machine and reuses the sessions they hold. Open Settings → Providers to see what's connected.
Claude
Kalauz reads claude auth status to detect your login and shows whether you're
Connected, your email, and your plan (for example Claude Max or Claude
Pro). If the CLI isn't found, you'll see CLI not found — install Claude Code
and make sure claude is on your PATH.
Not connected yet? Kalauz opens a terminal running claude auth login so you can
sign in; the status refreshes once you're done.
Codex
Codex auth is read from your Codex home (~/.codex/auth.json by default). Kalauz
shows whether you're signed in and via what — a ChatGPT plan (Plus, Pro, Team) or
an API key. To sign in, run codex login. The Codex models stay greyed out in
the picker until the codex CLI is installed.
Defaults
Set the default model and default effort new workspaces start with, and optionally a separate model/effort used for review tasks. Per-session overrides always win — see Agent modes.
Managed settings (for teams)
An administrator can drop a settings file (~/.kalauz/settings.json) to pin
choices across a team:
defaultModel— locks the model picker to an approved model.claudeExecutablePath— points Kalauz at a specificclaudebinary.enterpriseDataPrivacy— disables non-essential agent traffic.environmentVariables— injected into every agent and console process.
When a setting is managed, its control shows as locked in the UI. See Enterprise for the bigger picture.
What's next?
- Agent modes — choosing models and effort per session.
- Agent behavior — how those choices reach the CLI.