Documentation
LLM models
Models are configured per organization, on two levels: providers (the endpoints) and profiles (named configurations that agents reference).
No vendor lock-in
Any endpoint compatible with the OpenAI APIs can be a provider, even a private or in-house model.
Costs under control
A small, fast model where it is enough, a more capable one where it is needed.
Providers and profiles
| Level | What it contains |
|---|---|
| AI Provider | Name, protocol (Chat Completions or Responses), API base address and key, masked in the console |
| Model Profile | Profile name, provider, model name and call parameters (temperature, maximum tokens, reasoning level, web search for the Responses protocol) |
- The default model is used by agents that do not name one.
- Each agent can choose a different profile in its Model field.
- The organization's time zone (Europe/Rome if not set) governs clocks and scheduled tasks.
Related pages
How to create a new agent
Step by step: project, agent node, instructions, model, tools, channel, testing and publishing.
Multiple agents that collaborate
Delegating a task, handing over the conversation, splitting work into subtasks.
Security
Roles, groups, credentials, enterprise sign-in and least privilege: how to protect an agent system.
Have an agent to take to production?Let’s talk.
We show you TeraBrain NX on your case: systems to connect, access and controls.