Retell AI builds telephony voice agents (agents that make and receive real phone calls) with pay-as-you-go pricing from $0.07-0.31/minute for voice agents, plus component costs for voice infrastructure, TTS, the language model, and optional add-ons like PII removal. Pick it when you specifically need phone-network integration, not just browser-based voice. It bundles components with its own pricing now, beyond a pure bring-your-own-key model.
Other good fits
Starting small with pay-as-you-go pricing and $10 in free credits before committing to volume.
+ Pros
Read pros
Retell AI specializes in telephony voice agents: agents that place or receive actual phone calls over the phone network (PSTN), not just browser-based voice chat. It bundles the pieces a phone voice agent needs (voice infrastructure, speech-to-text, a language model, text-to-speech) with its own component pricing, plus optional add-ons like a knowledge base, safety guardrails and PII removal.
Pick it when…
Your agent needs to make or receive real phone calls, which needs telephony integration a browser-only voice API doesn't provide.
You want to start small with pay-as-you-go pricing and free starting credits before committing to a larger deployment.
You want add-ons like PII removal or safety guardrails available as configurable pieces rather than building them yourself.
− Cons
Read cons
Component pricing means your actual per-minute cost depends heavily on which model and which add-ons you choose; a cheap base rate can look very different once a knowledge base, guardrails and a premium LLM are all stacked on top. This isn't purely bring-your-own-key anymore in the way the catalogue caveat describes; check current documentation for exactly which pieces you're billed for directly versus need your own keys for, since that can change.
Pick something else when…
You don't need real phone-network integration.OpenAI Realtime or Pipecat handle browser-based voice without telephony-specific setup.
You want a more polished, managed conversational voice product with less component assembly.ElevenLabs Conversational AI is a comparable managed option to check pricing against.
You need a framework to assemble your own pipeline including telephony as one option among several transports.Pipecat supports SIP/PSTN alongside WebRTC if you'd rather build than use a packaged product.
Avoid if
Per-minute costs across voice infrastructure, TTS, LLM and add-ons need to stay predictable; component pricing here adds up in ways worth modeling before committing.
Workflow
Configure your voice agent's components (voice, language model, any add-ons) → get a phone number → set up your call flow and any knowledge base the agent should draw on → run test calls checking both the conversation quality and the actual telephony behaviour (hold, transfer, hang-up) → monitor per-minute costs against your usage once live.
Starting credits: $10 free. Enterprise pricing is custom and adds dedicated support, custom compliance agreements, uncapped concurrent calls and role-based access control. Running effort is mostly watching which model and add-on choices you make, since LLM cost alone ranges more than 200x depending on which model powers the conversation.
Restrictions
Check current documentation for exactly which components (voice infrastructure, TTS, LLM, add-ons) are billed directly by Retell versus need your own provider keys, since that split isn't purely bring-your-own-key the way the catalogue caveat describes and can change.
First project
Documentation-based suggestion — not run here.
Set up one test agent and make a real test call.
Sign up and claim your starting free credits.
Configure a simple voice agent with one clear purpose (for example, answering a basic FAQ) and get a phone number.
Call the number yourself and have a short test conversation.
Check the call transcript and cost breakdown afterward against what you expected.
Done when you've completed one real test call, verified the transcript accurately reflects the conversation, and understood which cost components (voice infrastructure, TTS, LLM, any add-ons) contributed to the total.
Links & evidence
Documentation-based · not hands-on · Sep 23, 2026
Rests on Retell AI's pricing page and homepage, checked 23 September 2026. No account was created and no call was placed for this review.
Assemble a phone-ready voice agent from your chosen speech and language services.
Handle telephony, routing, conversation state, and interruptions in one layer.
Use it for appointment, support, or outbound calling workflows.
Owner-requested editorial use of public official interface screenshots, 15 September 2026. Copyright retained by the original publisher; attribution preserved. No open licence or hands-on course execution is asserted. Source asset: https://mintcdn.com/retellai/wM4UcKxv3px1bKim/images/create-agent-dropdown.png. PNG conversion preserves the complete frame.