Creating a Voice Agent

Every voice agent is configured on its own page, in sections — Name and status, Voice & greeting, Instructions, Phone numbers and Call recording — with its calls listed below. This page walks through the create dialog and every field, explains what each one does, and points out the settings most teams get wrong on the first try.

Where Agents Live

Voice agents live at Brand → (your brand) → AI agent → Phone, on the Phone agents page. Click Create agent and the New phone agent dialog opens, where you can start from a prebuilt template, start from scratch, or use Draft with AI.

A voice agent is always tied to:

  • One brand — the brand is where its knowledge, its privacy masking, its company details and the Answer instructions of its AI agent all come from.
  • Zero or more phone numbers (the agent answers calls coming to these).
  • One voice + language combination (the spoken side of the conversation).

You can have several agents on one brand — for example a customer support agent on the main toll-free number and a separate after-hours agent on another. Each brand keeps its own set; agents are never shared between brands. For what a minute costs and what limits apply, see Pricing & limits.

The list shows each agent's name, its voice and language, and the number that rings it, with a status of Live, No number or Paused. An agent with nothing routed to it shows No number. An agent that has a number but whose brand has no knowledge yet shows "This brand has no knowledge yet — the agent can't answer questions." It still answers the phone, but it never makes up an answer from general knowledge, so add knowledge to the brand before you route real calls to it. Each row has a ⋯ menu with Edit, Pause or Resume, Duplicate and Delete, and a search box appears once you have six or more agents.

The New Phone Agent Dialog

The dialog has two steps.

  1. New phone agent — pick a starting point: Receptionist, Customer Support, Sales or After-hours FAQ (each has a play button so you can hear its voice), Start from scratch (an empty agent with a default voice, so you write the instructions yourself) or the Draft with AI row. Click Continue.
  2. Set up your phone agent — check the basics: Name, Voice, Language, Greeting (the fixed AI notice is shown above it), Instructions and Phone number (Not now, or one of your numbers that no agent uses yet). Tools, recording and call history live on the agent's own page. Click Create agent and that page opens.

If the number can't be attached, the agent is still created and a message asks you to pick the number on its page.

Name and Status

Defines who the agent is and whether it is answering calls. The section is headed with the agent's name, and under it a Manage knowledge link takes you to the brand's Knowledge page, where the agent's answers come from.

Name
text required
Only you and your team see it — callers never hear it. It shows in the agent list and in the Answered by column of the Phone numbers page. Pick something your team will recognize (for example "EN Receptionist · main toll-free").
Status
button
Shows Active or Paused, with a Pause or Resume button. Pausing takes effect at once and is not part of Save.

Voice & Greeting

Defines the language and voice, and how the agent opens a conversation.

Language
select
The language it greets and answers in. It controls the default greeting and the language every call starts in. See Voice & language for the 17 options in the list — and for the two other "language" settings that are easy to confuse with this one.
Voice
select required
One of 30 voices, shown as cards grouped into families, with a search box and Female / Male buttons. Press play to hear the AI notice and your greeting in that voice. The voice itself is multilingual — the same voice handles every supported language, so picking a voice never locks you into a language.
AI notice
text
Read-only. Every call opens with a fixed sentence in the agent's language — in English, "This call is answered by an AI assistant." — before your greeting. It is the disclosure the EU AI Act (Article 50) requires, so it can't be edited, removed or skipped, and callers can't interrupt it. It changes with the Language setting above.
Greeting
text
What the agent says right after the AI notice. Leave it empty to use the default greeting for the selected language (shown in the box as a hint). Keep it short and don't repeat that the agent is an AI — the notice already said so. If you have recording switched on, this is where your recording consent line goes.
Callers can interrupt the greeting
boolean
When on, speaking during your greeting stops it and the agent starts listening. Off by default — the greeting plays in full first. Either way the AI notice always plays to the end before your greeting begins.
Saving prepares the AI notice and your greeting as audio in the chosen voice and language, so the first seconds of a real call are instant. Changing the voice, language or greeting prepares that audio again automatically. The play button next to a voice previews the notice and the greeting together, exactly as a caller hears them. A preview you haven't played yet in this visit is a small paid voice generation, billed like the rest of your AI usage.

Instructions

Instructions
textarea required
What the agent does on a call: its role, tone and the steps it follows. Tell it who it represents, the personality and tone to use, what it should and shouldn't talk about, and any non-obvious rules ("never quote prices for enterprise plans, end the call and let the caller follow up by email"). The agent also follows the brand's company description and the Answer instructions of the brand's AI agent (without an AI agent, the brand's details), so you don't need to repeat them here. Unless When a caller asks for a person, below, says so, it never offers to take a message or promises a callback, even if instructions written earlier ask it to.
When a caller asks for a person
textarea
Optional, up to 2,000 characters. A call can't be transferred, so write what the agent should do instead when a caller asks for a person, wants a callback or wants to leave a message — for example "Offer a callback and ask what it's about." The agent says honestly that it can't connect the caller on this call, then follows your text. It stays on the line afterwards, it won't promise any time (not even "soon") unless your text names one, and the caller always leaves with a way forward: if your text gives none, the agent gives your company's contact details. Left empty, nothing changes: it says it can't connect the caller and gives your company's contact details.When the agent agrees something your team has to do, such as a callback, it tells your team, once per call. The conversation is marked Waiting for reply in the inbox, your team gets an alert with a short summary, and the same summary is added to the conversation as a private note (masked when the brand's privacy masking is on). The agent only tells the caller it passed the request on once that has gone through. If the caller's number is withheld and the summary has no way to reach them, the agent first asks how to reach them and reads it back. A text that only sends the caller somewhere else, such as to another number or an address, needs no alert. A conversation waiting for your team keeps waiting when that caller calls again and when a call ends, until someone resolves it.
When it has no answer
text
Read-only, not a setting. The agent says plainly that it doesn't have that information and never guesses. It gives your company's contact details from the brand's knowledge, or names your website if it finds none. A call can't be handed to a person, and the agent never offers to take a message or promises a callback, so there is nothing to choose here.
Write the instructions like you're onboarding a new hire, not like you're writing a chatbot prompt. The agent does best when given a clear role, a few hard rules, and explicit examples of when to stick strictly to what the brand's knowledge says.

Phone Numbers

Phone numbers
multi-select
The numbers that ring this agent: calls to these numbers ring this agent. A number can only be attached to one agent at a time; a number used by another agent is marked Assigned to that agent, and picking it here moves it over. You can do the same thing from Settings → Connections → Phone numbers — both write the same assignment. Under the field, Need another number? links to Buy one in Settings → Phone numbers.

Call Recording

Record calls
boolean
Saves a stereo MP3 of every call, with the caller on one channel and the agent on the other. Recordings are stored privately and capped at 100 per organization — the oldest are removed first when it is full. Switching it on shows a You are responsible for compliance warning, because recording rules differ by country and use case.
You don't choose an AI agent for a voice agent. A voice agent is a sibling of the brand's chat AI, not a child of it: both answer from the same brand knowledge. That is also why privacy masking and the company details come from the brand rather than from a bot. Your API tools and workflows come from the brand too: everything the brand has, including workflows that can run on a phone call, is offered to its voice agent, whether or not the brand has an AI agent.

Privacy Masking on Calls

Masking is switched on per brand, at Brand → (your brand) → Brand settings → Advanced → Privacy masking. It hides email addresses, phone numbers and card numbers.

One toggle, two behaviors — and on the phone it means the opposite of what it means in chat. In chat it strips personal details out of a message before it is sent to the AI provider. On a call it cannot do that: the caller's speech has already reached the AI in real time. What it redacts on a call is what we store — the saved transcript and the call's messages in the inbox. It is best-effort, never a guarantee: if the masking service is slow or unavailable, the text is stored unmasked and nothing records that it happened. It is not encryption, and calls recorded before you switched it on stay as they were.

Talk to It

The Talk to it button in the title bar of the agent's page lets you try the agent without a phone. A panel opens at the side with Start call, Hang up and, once the call is over, Call again. Allow your microphone and talk. What you and Yaplet say shows in the panel as it is spoken, along with what it looked up. The same call is also on the brand's Playground: switch its test area from Chat to Call.

  • It uses the last saved version of the agent, so unsaved changes are not included.
  • It works on a paused agent too. The agent must belong to a brand.
  • A test call stops after 5 minutes, and each person can make up to 10 an hour.
  • Use headphones, or Yaplet hears itself through your speakers.
  • Actions run for real: workflows and API tools really run.
  • If When a caller asks for a person is filled in and the agent agrees something for your team, the panel shows a Told your team: … line instead. Nothing is sent: no alert, and no conversation waits in the inbox.
  • It is charged to your workspace the same way as a real call, so it needs a paid plan, like phone calls.
  • It leaves no conversation in the inbox, and it doesn't use up the per-caller limits that apply to real callers.

Calls

Below the sections, a Calls table lists every call this agent has handled, real calls and test calls, defaulting to the last 30 days and filterable by date. Each row shows the caller, date, duration, cost and status — Ended, In progress, Failed, Session lost or Declined — daily limit — and lets you play the recording (if there is one) or open the conversation in the inbox. A test call shows the caller Test call and a Test badge, and has no conversation to open.

Generating an Agent with AI

The Draft with AI row, in the first step of the dialog, asks What should this agent do? — which calls it handles, who calls and what it should get done. Click Draft agent and Yaplet writes the name, the instructions and the greeting and picks a voice. It already knows this brand's company name, website and tone, so you don't repeat them, and the draft lands in the same details step for you to check ("Yaplet drafted this from your description. Check it, then create.") before you create the agent. If the draft fails, Try again takes you back to your description. If the brand has no AI agent yet, the panel says "This brand has no AI agent yet" and the Draft agent button stays disabled — start from a template instead.

Editing an Agent

Click any agent in the list to open its page, with the same sections. Changes are saved with Save changes in the title bar (Discard undoes them, and leaving with unsaved edits asks first). The spoken audio is prepared again whenever you change a field that affects it (voice, language or greeting). You can also Pause an agent from here: while it is paused, incoming calls are not routed to it, and Resume turns it back on.

Editing the instructions does not affect a call already in progress — the agent keeps the instructions it started with for the rest of that call. The new instructions take effect on the next call.

Duplicating an Agent

Use Duplicate in a row's ⋯ menu on the Phone agents page. It creates a copy named "… (copy)" with the same instructions, voice, language and greeting, but without a phone number — a number rings one agent only.

Deleting an Agent

Deleting an agent unassigns its phone numbers. Delete from a row's ⋯ menu on the Phone agents page, or from the ⋯ menu on the agent's own page (Delete agent): either way the confirmation tells you how many numbers become unassigned. Past call transcripts and recordings stay in the inbox. Deleting the brand deletes its voice agents too.

Unassigned is not the same as released. Those numbers stay on your organization and keep costing their monthly rental until the organization owner releases them at Settings → Connections → Phone numbers. See Phone numbers.

© 2026 Yaplet. All rights reserved.