Voice agent questions, answered
These answers describe the audentic website voice agent available today. Start with the quickstart if you are ready to build one.
What is audentic?
audentic is a managed platform for creating and running voice agents on websites. You describe the agent's job, add business information, choose a voice, test a conversation, and publish a widget. audentic handles the voice connection and model access. See the platform overview.
Is audentic a free service?
audentic offers Free to get started, with platform-provided model access and no platform subscription fee. Free access is subject to platform limits. Subscription plans are coming soon. Compare plans or manage your account in Plan & billing.
Where can I see usage and costs?
The business dashboard and Usage & costs show conversation counts, completed connection time, and activity by agent. Current voice conversations do not yet have verified costs; unavailable costs are not zero-cost usage. Historical recorded costs are labeled separately. See usage and billing.
Do I need to write code or bring an OpenAI API key?
No. Free lets you configure, test, and publish an agent without writing code or adding an OpenAI key, subject to platform limits. You need a place on your website where you can paste the generated snippet. A developer can use the React package instead.
Can I create an agent by talking to it?
Yes. Sign in, choose Create agent, then Use voice or type instead and Talk to the builder. Describe the agent's job, add facts, and ask for changes. The builder saves a draft as you talk. End the call and wait for the final changes to sync before testing, editing, and publishing. See the quickstart.
Do visitors need an audentic account?
No. Visitors can use a published voice agent on a website whose origin you approved. The agent owner signs in to configure it and review conversations.
Does the microphone start automatically?
No. The visitor starts the voice connection and grants browser microphone permission. The Greet visitors when they connect setting only controls whether the agent speaks first after that connection starts.
Can visitors use it without a microphone?
Yes. Choose Type instead in the widget. It still needs an internet connection and available conversation allowance. You can also save a contact page or email in the builder as another way for visitors to reach your team.
Can a paused conversation continue?
Choose Resume conversation when the same widget offers it after a pause or disconnection. It starts a new connection using saved context, without repeating confirmed requests. The latest words may be missing if they were not saved. Reloading the page or closing the widget can remove the resume option.
Will it work on my website?
It can work on a website that allows custom HTML and scripts, microphone access, and a secure HTTPS page. Local development can use localhost. Add the site's exact origin, including its port when relevant, to the agent's allowed website list, then test a real conversation on the published page. See website setup and browser requirements.
Can the agent answer from my documents or website?
Yes. Start with one public HTTPS page to create a private draft, then review the imported facts and source quotes before publishing. For an existing agent, you can paste business facts, add a note, import text from one public HTTPS webpage, or upload a supported file such as a PDF. Review extracted text before saving it. audentic does not automatically crawl your entire website. See knowledge sources and limits.
What happens when a visitor asks for follow-up?
If you enable Take a message, the agent can prepare a request with the visitor's contact details and message. The visitor reviews and confirms it on screen before audentic saves it in your Inbox. Email alerts and HubSpot capture are separate, optional connections. Enable them under Follow-up connections when available; otherwise requests stay in Inbox. Text notifications are not available. See follow-up requests.
Can the agent show choices or collect feedback in a form?
Yes. Enable Show helpful steps on screen. During a conversation, the agent can create choices or a form with fields suited to what the visitor said, then revise the form as needs change. The visitor reviews and edits an owner-directed form, then explicitly presses Send to submit it to your Inbox. The form can be completed shortly after the voice call ends. If email alerts are available and enabled for your account, audentic can notify you with a link to Inbox.
Can the agent update what visitors see while they talk?
Yes. With on-screen steps enabled, it can revise a form or view inside the audentic widget. A separate owner setting allows custom interactive screens. Their HTML, CSS, and JavaScript run in an isolated frame. The agent cannot edit the rest of your website, and owner-directed form entries are sent only after the visitor presses Send.
Can it fill forms or book appointments for visitors?
It can create and revise forms inside the audentic widget. It cannot fill arbitrary forms elsewhere on your website, control the browser, or book appointments in external systems. Those broader actions are exploratory roadmap work.
Does audentic save audio recordings?
The app saves conversation checkpoints and a final transcript; it does not store a call audio recording. A closed tab or lost connection can still leave the latest transcript incomplete. See conversation history.
Can ChatGPT or Claude Code create and publish agents for me?
Today, a coding assistant can add an existing, published audentic agent to your website code. audentic does not yet provide a supported public management API, CLI, or MCP server for agents to create or publish agents in your account. Follow the coding-agent integration guide; owner-authorized management is planned.