Every agent answers with an AI model working behind it: its brain. You do not have to be an expert to choose one. Three plain ideas cover it:
- The provider is the AI company.
- The model is one specific brain from that company.
- The reasoning level is how deeply the agent thinks before it answers.
Your agent keeps its name, personality and knowledge whichever brain you pick, and you can change the brain at any time.
Where the brain is set
Open your agents list and press the wand button titled Unified Editor on the agent. Choose the second row of tabs, Channels, then Text Chat. Under AI Model Configuration, the Primary LLM menu is the brain that answers in chat; the provider follows from the model you pick. Press Save, then Chat at the top of the editor to hear the difference for yourself.
The providers NetShow works with are GPT from OpenAI, Claude from Anthropic, Gemini from Google and Grok from xAI. There is no provider that is right for every agent. Pick for the job it does most often, then judge it by its answers, not by the name.
How deeply it thinks
Where your agent's settings show a Reasoning Level picker, it offers four levels, from the fastest to the most thorough:
| Level | What it is for |
|---|---|
| Quick | Fast answers for simple questions. Lowest cost. |
| Standard | Balanced intelligence for everyday tasks. |
| Deep Think | Extended reasoning for complex problems. |
| Maximum Intelligence | Absolute maximum reasoning for critical decisions. |
Each step up thinks longer and costs more, and Maximum Intelligence costs the most. The picker's own advice is the right starting point: start with Standard for everyday tasks, and switch to Deep Think when the answer really matters. Keep Maximum Intelligence for the questions where a better answer is worth the wait.
You do not have to learn model names. Each level already knows which model of each provider fits it, and NetShow moves that choice forward as providers release newer models. If the agent feels too shallow, raise the level. If it feels slow or costly for routine questions, lower it.
The other settings on Text Chat
- Temperature sets how varied the answers are. Lower is steadier and more predictable; higher is more varied. The middle suits most agents.
- Max Tokens caps how long one answer can be.
- System Prompt Override replaces the agent's own instructions for chat. Leave it blank to keep using its personality and soul files; that is almost always what you want.
- Communication Protocol is technical. Leave it as it is.
If you chose wrong
Nothing is lost. Change the brain or the level, save, and ask the same question again in Chat. A good test is three questions: one from its everyday job, one that needs careful thought, and one it cannot know, to see how it handles not knowing. Its personality lives in the Personality tab and its facts in the Knowledge tab, so neither moves when the brain does. See Shape what your agent knows and says.
What this does not do
This guide does not quote an amount or promise which exact model sits behind a level today: the editor shows the model it picked. It does not ask you for a key. Changing the brain changes how the agent thinks, not what it is allowed to do; its limits are set elsewhere in the editor.
