Providers, models, and capabilities
Understand provider identity, model selection, and the settings that depend on each.
Updated 2026-09-15
Farero runs Codex, Claude Code, and OpenCode through their provider interfaces. A provider is the software that runs the agent session. A model is the selected intelligence behind that session. Their names answer different questions.
Where to look
| Question | Where to check |
|---|---|
Is the provider installed and signed in? | Settings → Providers |
Which model is selected? | The conversation's model picker |
Can this selection use Plan, effort, or Fast? | Its available conversation controls |
Which configuration will a helper use? | The selected agent definition |
Can it generate images or operate a browser? | The configured provider's actual tools and capabilities |
Provider settings include permission, sandbox, approval, extra-instruction, and environment choices. A model change does not make these boundaries disappear.
Auto and explicit choices
Auto routes a new prompt through the configured classifier. An explicit choice identifies the model directly. Managed Workflows use their frozen role settings independently of ordinary chat controls.
Use the current picker and configuration as your availability reference. A named model in an example may be absent from your account or provider. Cursor is outside Farero's supported provider set.
For setup, return to Connect a coding provider.