Model pickers that know what each coding CLI offers your account

The agent picker lists the models your account can use, read from Cursor Agent, Antigravity, Grok Build, Pi, Oh My Pi, Kiro CLI, Devin, Kimi Code and Qwen Code, as it already did for Codex and OpenCode. Lists are kept for a day so the picker opens quickly, and Refresh asks the CLI again.

Effort follows the model. Codex shows each model's own reasoning levels and its default, and Claude Code, OpenCode, Grok Build, Pi, Oh My Pi, Kiro CLI, Hermes Agent and Aider get an Effort setting.

Each agent remembers its own model and effort, so switching CLIs no longer hands one CLI's model to another. A saved model the CLI no longer offers is flagged in the picker before the next turn fails.

Claude Code adds Fable to its model list.

Long model lists get a filter box and are grouped by provider. The picker says whether a list was just read, cached or built in, and shows why when a CLI could not list its models. Hermes Agent, which cannot list models, takes a typed provider/model id with an example.

A model your account cannot use now shows a "cannot use this model" card with a Model badge. When no model was picked, the card says the CLI's own default model is the cause.

Expired logins, wrong API keys, usage limits, and the errors Qwen Code and Hermes Agent print as replies now show as error cards instead of empty answers. When OpenCode keeps retrying a failing model, a status line says so instead of the run looking stuck. A failure is reported once, not twice.

Kiro CLI runs under every permission preset and can use the browser tools. Codex and Cursor Agent no longer refuse browser tool calls in chat. Antigravity and Kimi Code no longer offer browser tools, which they could not receive.

Each agent shows its own logo in place of a two-letter tile. When Pi, Oh My Pi or Hermes Agent is not installed, the picker shows its install command.