---
title: "Agent Settings"
description: "Configure your agent's name and role, conversation history, credit limit, AI model, instructions, attachments, and training."
---

In your agent, select **Settings**. The page is split into four sections: **General**, **AI**, **Training**, and **Security**. Use the section links to jump between them.

## General

### Agent details

- **Name**: what the agent calls itself, shown to customers.
- **Role**: what the agent does, such as "Customer Support". Optional.
- **Agent ID**: the agent's unique ID, used in the [web widget script](/docs/deployments/web-widget/installation). Select the copy icon to copy it.
- **Knowledge size**: how much knowledge the agent has.

### Conversation history

**Conversation persistence** decides whether the web widget and Public Chat remember a visitor's conversation on their browser:

| Option | What happens |
| --- | --- |
| **Session only** | The conversation resets when the page reloads. |
| **Remember returning visitors** | The conversation is kept on that browser, so visitors can pick up where they left off. |

WhatsApp conversations aren't affected.

### Credits limit

Cap how many of your team's [credits](/docs/billing/understanding-credits) this agent can use each billing cycle. It stops one busy agent from using up credits your other agents need.

1. **Turn it on**

    Switch on **Set credits limit on agent**.

2. **Set the limit**

    Enter the number of **Credits per cycle**. The card shows how many the agent has **Used this cycle** and how many remain at the limit.

When the limit is off, the agent shares the team's credits with every other agent.

## AI

- **AI model**: the model that writes the agent's replies. See [AI Models](/docs/agents/ai-models).
- **Base instructions**: how the agent behaves. See [Agent Instructions](/docs/agents/instructions).
- **Attachments**: lets customers send images, PDFs, and voice messages. Off by default. See [Attachments and Voice Messages](/docs/agents/attachments).

## Training

- **Last trained at**: when the agent's knowledge was last trained.
- **Auto-retrain**: retrain automatically every 24 hours to pick up changes. See [Retraining Knowledge](/docs/knowledge/retraining).

## Security

The rate limit and user identification settings are covered in [Agent Security](/docs/agents/security).

## Danger zone

- **Delete conversations**: permanently removes all of this agent's conversations. The agent, its knowledge, and its settings are kept.
- **Delete Agent**: permanently removes the agent and all its data. See [Creating and Managing Agents](/docs/agents/creating-and-managing-agents#delete-an-agent).

Neither action can be undone.