> ## Documentation Index
> Fetch the complete documentation index at: https://mind-e.co/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Customizing Agent Behavior

> Fine-tune your agent's responses with system prompts, model selection, temperature, and query expansion.

After creating an agent, you can adjust its behavior through the **Settings** tab. These settings control how your agent interprets questions and generates responses.

## System prompt

The system prompt is the most important setting. It tells the agent who it is, how to behave, and what rules to follow.

### Writing a good system prompt

* **Define the role** — tell the agent who it is and what it does
* **Set boundaries** — specify what the agent should and shouldn't do
* **Define the tone** — formal, friendly, concise, detailed
* **Add rules** — specific instructions for handling edge cases

**Example for customer support:**

```text theme={null}
You are a customer support assistant for Acme Corp. Your role is to help
customers with questions about our products and services.

Rules:
- Answer based on the provided context only
- If you don't know the answer, say "I don't have that information" and
  suggest contacting support@acme.com
- Be friendly and concise
- Never make up product features or pricing
- For billing issues, always direct the user to the billing team
```

**Example for education:**

```text theme={null}
You are a study assistant for an introductory biology course. Help students
understand concepts from the course materials.

Rules:
- Explain concepts in simple terms with examples
- When citing from course materials, reference the source
- Encourage students to think critically rather than giving direct answers
- If a question is outside the course scope, let the student know
```

<Tip>
  Start simple and iterate. Test your agent in the playground, see where it falls short, and refine the prompt.
</Tip>

## Model selection

Choose the AI model your agent uses. Different models have different capabilities and [credit costs](/core-concepts#credits):

| Tier     | Models                                             | Credits/message | Best for                            |
| -------- | -------------------------------------------------- | --------------- | ----------------------------------- |
| Mini     | GPT-4o Mini, Gemini Flash, Claude Haiku            | 1               | High-volume, simple queries         |
| Standard | GPT-4o, GPT-4.1, Gemini Pro, Claude Sonnet, Grok 2 | 2               | General-purpose (recommended)       |
| Premium  | GPT-5.2, Claude Opus, Grok 3                       | 3               | Complex reasoning, detailed answers |

GPT-4o (standard tier) is the default and offers a good balance of quality and cost for most use cases.

## Temperature

Temperature controls how creative or consistent the agent's responses are. The range is 0.0 to 2.0.

* **Low (0.0–0.5)** — more consistent, factual, predictable. Good for support bots and factual Q\&A.
* **Medium (0.5–1.0)** — balanced. Good for general-purpose agents. The default is **0.7**.
* **High (1.0–2.0)** — more creative, varied, unpredictable. Good for brainstorming or creative writing.

<Note>
  Some models have a fixed temperature that cannot be adjusted. The setting will appear locked for these models.
</Note>

## Query expansion

When enabled, Mind-e automatically expands short or vague user queries with additional keywords before searching the knowledge base. This improves retrieval quality for brief questions.

**Example:** A user asks "pricing" → the system expands it to something like "pricing plans costs subscription tiers features" before searching.

Query expansion is enabled by default. You can toggle it in the agent settings. It costs an additional **2 credits** per message when it triggers.

<Tip>
  Keep query expansion enabled if your users tend to ask short questions. Disable it if your users write detailed queries or if you want to save credits.
</Tip>

## Saving changes

After adjusting any settings, click **Save** to apply the changes. Settings take effect immediately for new conversations.