Introduction to Retell

​

How does a voice AI agent work?

A voice agent turns a phone call or web session into a conversation by connecting four layers:

1. Telephony. The call arrives over the phone network, over a SIP trunk you already own, or starts in a browser or app.
2. Speech to text. The caller’s audio is transcribed while they talk.
3. Language model. The transcript goes to an LLM that decides what to say next, using the agent’s prompt, tools, and knowledge base.
4. Text to speech. The reply is turned back into audio and streamed to the caller with minimal delay.

Wiring those four together yourself is the easy part. Keeping the conversation natural is not: the agent has to know when the caller has finished speaking, recover when it is interrupted, hold up through background noise, and fall back cleanly when a provider degrades mid-call. Retell runs all four layers and handles that orchestration for you, so you build an agent as a single prompt or a conversation flow instead of assembling providers. See orchestration overview for how the layers fit together, and speech recognition, LLM options, and voices for the choices at each one.

​

Key capabilities

• A single prompt agent to start, or node-based conversation flows when you need tighter control over execution order (how to choose)
• Voice and chat agents from one agent definition
• A knowledge base your agent answers from, filled by crawling a site or uploading documents
• Call center handling: warm transfer, IVR navigation, and voicemail detection
• Post-call extraction for success and sentiment scoring, plus custom fields synced to your CRM
• Integrations for CRM, helpdesk, calendar, and file storage, connected once and available to every agent

​

Popular use cases

Lead qualification. Screen inbound leads on the phone, ask the qualifying questions, and book or route only the ones that fit. Start with a single prompt agent. Customer support. Answer common questions, look up order or account details in your systems, and hand the complex cases to a person. Start with a single prompt agent. AI receptionist and appointment scheduling. Answer after hours, check availability, and book, reschedule, or cancel appointments. Start with a single prompt agent. Patient scheduling and intake. Handle the structure healthcare calls need, including intake questions and confirmations. Consider a conversation flow agent if compliance requires a fixed order of steps. Dispatch and field service. Take service requests, capture the job details, and route them to the right technician or crew. Start with a single prompt agent. Agencies and resellers. Build agents for multiple client accounts from one platform, each with its own numbers, prompts, and analytics. Either agent type.

​

Build

Retell agents use a single prompt or a conversation flow, with call handling, a knowledge base, and integrations built in. We recommend starting with a single prompt; see Choose an agent type for when conversation flow is useful. The build overview covers the rest of the configuration.

​

Test

Test in text or real audio, manually or automatically, before an agent ever takes a real call. See the testing overview to compare every method.

​

Deploy

​

Monitor

​

Data

​

Developer friendly

Build entirely from the dashboard with no code, or go fully programmatic with the API, SDKs, and an MCP server.

​

Next up

Ready to dive in? Start here.

​

FAQ