<!--
  Translated by the docs agent from content-src/es/guides/proveedor-ia-propio/04-modelos-disponibles-acotados-a-tu-proveedor.md
  (source_hash 7a3ab2bfe3c2). If that Spanish source_hash changes, re-check this translation.
-->

With your own key active, your agents' model selector (Manual 23) is **scoped to your provider's models** — you can never mix, for example, your OpenAI key with a Claude model. If you have an agent in **Automatic** mode, it picks among those same scoped models.

> **Important — An agent pinned to another provider becomes "unavailable", with a reason**
>
> If one of your agents has its model **pinned by hand** to a provider other than your own key's, that agent stops responding while the situation isn't resolved — the platform **never silently changes the pin**, nor does it serve conversations with another provider behind your back. The visitor sees an honest notice ("right now we can't start new conversations on this channel; our team will get in touch with you shortly"), never a technical error or a half-finished reply. If the channel is **WhatsApp or Telegram**, the customer's message **is kept** in your conversations inbox — it isn't lost, it stays waiting for you to resolve it. You'll also get a notice (bell and, depending on your preferences, email) so you notice it as soon as possible.
>
> **How to fix it:** go to **AI Agents** (Manual 23) and change that agent's model to one from your own key's provider — or set it to **Automatic** so the platform picks on its own among the models already scoped to your provider.

If your provider doesn't have any model activated for your account (an exceptional situation, usually resolved by support), conversations on that channel stop with the same honest notice above until at least one model is available.