Skip to content

Models

Model and reasoning-effort selection is configured per agent, not globally.

AgentDock doesn't have a single global model picker — model selection is a per-agent setting, since each CLI exposes its own model catalogue and options.

Per-agent model selection

  • Claude Code and Codex support live model switching where the underlying CLI/SDK supports it — you can change models mid-session without starting a new one.
  • Codex's model list is fetched live from the CLI's own app-server, so it reflects whatever your installed Codex CLI actually offers.
  • Antigravity, being a terminal/PTY-based integration rather than an SDK integration, does not expose the same live model-switching behavior as the other two.

Reasoning effort

Each agent's configuration also includes a reasoning-effort field alongside the model selection, where the underlying CLI supports tuning it.

Model and reasoning-effort settings live in Settings → Agents — see that page for the exact controls, and Agent configuration for how detection and executable paths work.