"Models" and "Plans" answer different questions, even though they live in the same **AI** tab of your
console (Manual 40): a **plan** is how much and what you charge a customer; a **model** is **which specific
AI engine** (from OpenAI, Claude/Anthropic, Gemini or DeepSeek) that customer's agent runs on underneath.
This page explains the reasoning behind the model system; the full procedure — tab by tab — is in
**[Your plan catalog and AI](/docs/en/guides/planes-ia/)**, section 4.

## A closed catalog, never free text

Neither you nor your customers ever type a model's name by hand into any form on the platform: every model
selector — yours in **Models to use**, your customers' in their own agent — is populated from a **master
catalog** managed by Aymar Agents's team, curated and with a real, known price. This avoids the typical
problem of a free-text field for the model: misspelled names, discontinued models that keep "working" in an
old form, or a margin that breaks because nobody knows what was actually typed really costs.

## Models to use: your own activation

On top of that master catalog, you decide — provider by provider — which ones to turn on for your account,
in the **Models to use** tab under **AI**. Only what you turn on there shows up afterward everywhere else:
in your **Keys**, in your **Prices**, in your **Negotiated** prices, and in the model selector inside each
of your customers' agents. When you're signed up as a partner, your account starts with the entire master
catalog already on — nobody starts with an empty screen; if a model doesn't interest you, you turn it off
whenever you want, and if you're missing one that isn't in the catalog, the panel itself lets you request
it from support without leaving the screen.

## With your key, or with Aymar Agents's key

For every model you turn on, your customers use it in one of these two ways — and the screen always makes
clear which one applies in each case:

- **With your own key** (BYOK — "bring your own key"): you add your own API key for that provider (Manual
  40, section 4.2) and your customers' usage runs on your own account directly with the provider, outside
  the platform.
- **With Aymar Agents's key**: if you don't have your own key for that provider, your customers can still
  use it — usage is debited from your **Wallet** at your own prices (Manual 39). It's the friction-free
  route to offer a model without having to manage the relationship with each AI provider yourself.

You can mix both approaches within the same account: some providers with your own key, others with
Aymar Agents's, whichever suits you.

## How it flows through to your business with customers

Once you have models turned on and, if you want, your own keys, the rest of the cycle lives in the
**Prices** and **Negotiated** tabs of the same **AI** screen: you set how much you charge your customers
per model (a general list, or a negotiated price for a specific customer), and the **Margin** tab always
compares what you bill against what Aymar Agents debits you — never the AI provider's real cost, only
your own margin. The full procedure for each tab is in **Your plan catalog and AI**, sections 4.2 to 4.6.