---
title: "Create a voice agent"
description: "Describe your agent in plain language and let Speak AI draft it, start from a ready-made template, or build one from scratch."
---

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

# Create a voice agent

A voice agent is a configured AI persona: a name, a personality, instructions for how it should
behave, and a voice. Creating one is the first step before you can test it, share it, or put it to
work, and Speak AI gives you three ways to get from nothing to a working agent.


## Who can create an agent

Only account owners and admins can create voice agents. If you don't have that role, **Your
agents** shows the agents that already exist, but not the tools to build a new one.

## Get to the create screen

Open **Your agents** in the sidebar. If your team already has agents, select **New agent** in the
top right to open the create screen. If this is the first agent on the account, that create
screen's prompt box and template gallery appear directly on the page instead, under **Describe
the agent you want**. Either way, you land on the same prompt box with a template gallery below
it.

## Describe your agent in plain language

The prompt box is the flow Speak AI leads with: it sits above the template gallery. Type what the
agent is for, for example a customer support agent for a SaaS product, or a scheduling assistant
for a clinic. The hint under the box spells out what makes for a better result: include what the
agent does, your website, your brand's tone, and who it's talking to.

Select **Generate** (or press Cmd/Ctrl+Enter) and Speak AI drafts a name, a personality,
instructions, and a welcome message from your description. If you already have instructions
written, select **Add your own instructions** first and paste them in; they're folded in
alongside your prompt instead of replacing it.

When your prompt doesn't give Speak AI enough to work with, it asks a single follow-up question
rather than guessing, for example about who the agent is for or what it should never do. Answer it
and generation continues from there.

## Review the draft before it's saved

Once generation finishes, you get an **Agent preview**: the name, personality, instructions,
welcome message, and any conversation starters it wrote, laid out so you can check them before
anything is final. Each field has its own **Edit** control to change the text by hand, and its own
**Generate** control to have Speak AI rewrite just that field, either as-is or with a short note on
what to change first. When you're happy with it, select **Save changes** and the agent opens in
the full editor for anything else you want to configure.

## Start from a template

Below the prompt box, **Or start from a template** opens a gallery of ready-made agents grouped by
category, including Support, Sales, Research, and Meetings. Each card names the template,
describes what it's built for, and is labelled **Creates this agent now**: unlike the
plain-language path, there's no draft or preview step. Selecting a card creates the agent
immediately with that template's personality, instructions, and voice already filled in, then
opens it in the full editor so you can adjust anything before you use it.

## Start from scratch

If neither a description nor a template fits, select **Start from scratch** at the bottom of the
template gallery. It creates an agent named "New Agent" with generic default personality,
instructions, voice, and welcome message already filled in, then takes you straight to the full
editor so you can replace them with your own.

## Related pages

- [Create and manage Speak AI voice agents](/api/voice-agents/agents/) for the Agents and
  Generation API endpoints behind this screen, including the exact request and response shape for
  prompt-based creation.

Source: https://docs.speakai.co/help/voice-agents/create/index.mdx