Help center

Your guide to better voice experiences.

Find practical answers for building, launching, and managing reliable AI voice experiences.

Browse help

Find the right starting point.

Getting started

Set up your workspace, create an agent, and prepare your first test call.

5 guides

Voice agents

Configure instructions, voices, languages, tools, and conversation behaviour.

8 guides

Calls & conversations

Understand phone numbers, transfers, recordings, transcripts, and outcomes.

7 guides

Integrations & API

Connect business tools, configure webhooks, and troubleshoot API requests.

9 guides

Analytics

Review transcripts, call outcomes, trends, quality signals, and exports.

6 guides

Workspace & billing

Manage teammates, permissions, usage, invoices, and subscription settings.

6 guides

Common questions

Quick answers, clearly explained.

Start here for the questions teams ask most often while designing and operating AI voice agents.

How do I create my first AI voice agent?

Start by defining the agent's purpose, caller goals, and boundaries. Choose a voice and language, add the information or tools it needs, then test common paths and edge cases before publishing.

Can a voice agent transfer a call to a person?

Yes. You can define handoff conditions for complex, sensitive, or high-value calls. A good handoff includes the caller's intent, collected details, transcript context, and actions already attempted.

How should integrations access business data?

Use the narrowest permissions possible, validate every request, and require confirmation before sensitive actions. Keep tool activity observable so your team can review failures and unusual behaviour.

Where can I review transcripts and call outcomes?

Conversation analytics should provide transcripts, detected intent, call outcome, transfer details, tool activity, and quality signals. Access may depend on your workspace role and retention settings.

What should I test before launching an agent?

Test normal requests as well as silence, interruptions, corrections, noisy audio, unexpected questions, tool failures, and human escalation. Confirm that every call ends with a clear outcome.