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

# Agents

> The voice or chat agent you're testing, and how you connect it.

An **agent** in Skrut is the AI system under test — the thing Skrut will have conversations with and grade.

## Voice agents

Connect via **LiveKit**. Skrut joins the room as a caller, speaks test utterances, listens to your agent's responses, and grades each turn for accuracy, latency, and policy compliance.

## Chat / RAG agents

Connect via your **HTTP endpoint**. Skrut sends messages as JSON, reads the reply, and continues the conversation across multiple turns. Works with any framework — LangChain, LlamaIndex, custom — as long as you expose a `/chat` or equivalent endpoint.

## Connection settings

| Field    | Description                                            |
| -------- | ------------------------------------------------------ |
| Name     | A label for this agent in your dashboard               |
| Type     | `voice` or `chat`                                      |
| Endpoint | LiveKit room URL (voice) or HTTPS endpoint (chat)      |
| Auth     | Optional bearer token passed in `Authorization` header |