Updated: 2026-07-24
The agent brain is the language model (LLM) that phrases the replies. In Corpis it is provider-agnostic: the platform talks to the model through a compatible interface, and the model itself is chosen in the agent settings. Character, rules and body are stored apart from the brain, so you can switch the model without recreating the agent.
A Corpis agent has two layers. The body and character are what you define: the name, manner of speaking, behaviour rules, greeting, knowledge and appearance. The brain is a rented language model that, based on this character and the current conversation, phrases the actual reply.
These layers are independent. The character describes WHAT the agent should be; the brain handles HOW each line is worded. That is why the brain can be treated as a swappable part, not as a piece of the agent's identity.
The model is set in the agent settings — the same place as the name, character and rules. The platform reaches the LLM through a compatible interface, so it is not tied to a single vendor: any model that speaks this interface can fit.
The chosen brain works across all of the agent's channels at once — the on-platform chat, the connected Telegram bot, the web widget and the public page. There is no need to configure the model per channel.
Switching the model does not touch what you have already built. Character, rules, greeting, knowledge base, memory and body are stored apart from the brain — when you switch models they stay in place, and the agent does not need to be recreated.
Only the wording changes: different models build phrases differently and hold long context differently. The easiest way to see the difference is the preview — ask the same question before and after the switch.
The default brain is set by the platform — you do not have to configure a separate model for an agent. This is a default, not a hard dependency: the provider-agnostic interface exists precisely so the model can be replaced when a better fit for the task or budget appears.
Brain usage is measured in credits regardless of which model is selected; limits and your remaining balance are shown in the billing section.
The brain is provider-agnostic: the platform talks to the LLM through a compatible interface, and the model is chosen in the agent settings. The default brain is set by the platform, but the interface is not tied to a single vendor.
No. Character, rules, greeting, knowledge and body are stored apart from the brain. When you switch models they stay in place — only the wording changes, not the agent itself.
No. The body and character are defined by you and do not depend on the brain. The model only handles how a given line is phrased, so it can be switched without touching the agent's appearance or personality.