> ## Documentation Index
> Fetch the complete documentation index at: https://docs.vivi.bot/llms.txt
> Use this file to discover all available pages before exploring further.
# Overview
**Agents** are the core of the VIVI platform. An agent is your virtual team member, interacting with guests, staff, or systems on your behalf. Agents combine components — like knowledge bases and prompts — into one system that delivers helpful, accurate, and efficient support. At a minimum, every agent needs a **name**, **prompt**, **model**, and **channel** to go live. Once configured, use **Test Agent** or **Evaluations** to validate responses before deployment.
Unlike traditional decision trees or scripted bots, VIVI agents are **dynamic** — they reason, act, and adapt to context. Agents can be duplicated, customized, or scaled as your organization evolves.
***
## Building Agents
You can customize every detail of an agent to match your organization's standards, tone, and goals.
### Core Requirements
A clear name and description that reflects the agent's purpose.
Defines the agent's role, tone, and rules of engagement.
The large language model that powers the agent.
Where the agent will be active — web chat, WhatsApp, API, and more.
### Optional Enhancements
Connect internal sources of truth so responses stay grounded in your approved information.
Connect MCP servers, APIs, or even other agents for advanced actions.
Attach reusable, structured instructions to this agent's prompt.
Add custom-built skills to let the agent run code locally.
Let the agent reply in both voice and text, instead of text alone.
Include numbered references to knowledge base sources in responses.
Reply with a typed JSON object instead of free-form text on /invoke and /batch — useful for API-driven workflows.
Enable public web search so the agent can pull in information beyond its knowledge bases.
Set how many days this agent keeps conversation history before it's automatically deleted. Defaults to 90 days, and must be between 30 and 365 days.
### Before You Launch
Once your agent is configured, verify it before exposing it to guests or staff. VIVI gives you two ways to do this:
* **Test Agent** — Chat with your agent directly to check tone, accuracy, and behavior in real time. This is the fastest way to catch issues while you're still tuning a prompt or knowledge base.
* **Evaluations** — Run your agent against a saved dataset of questions to measure accuracy and groundedness at scale, useful once you're validating a larger set of scenarios or comparing changes over time. See [Evaluations](/documentation/features/evaluations/runs) for setup details.
We recommend Test Agent for quick iteration, and Evaluations before any major launch or update.
***
## Best Practices
* **Use descriptive names** so team members can quickly identify each agent's role.
* **Enable citations** when your agent draws from knowledge bases — it builds user trust and makes responses easier to verify.
* **Use structured output only for API-driven workflows** where a downstream system needs to parse the agent's response as JSON.
* **Set conversation retention deliberately** — shorter windows reduce how much guest data you store, while longer windows help with support follow-ups and evaluations.
* **Test thoroughly** after configuring each component to ensure accuracy and reliability.