> For clean Markdown of any page, append `.md` to the page URL.
> For a complete documentation index, see https://docs.sarvam.ai/llms.txt.
> For full documentation content in one file, see https://docs.sarvam.ai/llms-full.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.sarvam.ai/_mcp/server.

# Overview

This section covers everything about a voice agent: what a conversational agent is, how it's architected, its core primitives, and how you design, write, build, and improve one. Use the links below to jump to any part.

## 1. What is a conversational agent?

Start here for the mental model.

#### [Introduction](/conversations/build/concepts/introduction)

Voice vs chat, channels, and what Voice Agents covers today.

#### [Models](/conversations/build/models)

The constellation of models behind every turn.

#### [The harness](/conversations/build/concepts/harness)

Context, tools, and the agentic loop that make models useful.

## 2. Core primitives & concepts

The building blocks of an agent.

| Concept                         | Guide                                                                         |
| ------------------------------- | ----------------------------------------------------------------------------- |
| The agent and its parts         | [Agent](/conversations/build/agent/overview)                                  |
| How the agent behaves           | [Instruction](/conversations/build/system-prompt)                             |
| Per-call data                   | [Variables & personalization](/conversations/build/variables-personalization) |
| Language & how it sounds        | [Speakers & voice](/conversations/build/voice-language)                       |
| Acting on the world             | [Tools](/conversations/build/tools)                                           |
| Context at call boundaries      | [On-start & on-end hooks](/conversations/build/on-start-on-end-hooks)         |
| Grounding in your content       | [Knowledge base](/conversations/build/knowledge-base)                         |
| Getting the prompt right        | [Prompt Practices](/conversations/build/single-state-agents)                  |
| Turn-taking, timeouts, language | [Configuring settings](/conversations/build/conversation-settings)            |

## 3. Building & improving agents

The workflow from first draft to a deployed agent you keep improving.

#### [The build loop](/conversations/build/process)

Draft → test → deploy → measure → improve.

#### [Evaluation criteria](/conversations/build/process#goals)

Define what "success" means and measure against it.

## 4. Navigating the workspace & Genie

#### [Navigating the build interface](/conversations/build/navigating-workspace)

Versions, canvas, settings, and the things to be careful about.

#### [Genie](/conversations/build/genie)

The AI assistant that helps you build and review agents.

## 5. Advanced concepts

When a single prompt isn't enough.

#### [Multi-state agents](/conversations/build/states-conversation-flow)

Split a conversation into phases with gated tools and transitions.