Rexa.ai
Get started
FAQ

Questions, answered directly.

48 answers on pricing, setup, telephony, calling, transfers, recordings, integrations, latency, languages, security and compliance.

Jump to a category

Pricing and billing

How much does Rexa.ai cost?

Rexa.ai costs 10¢ per minute of conversation. On the default setup that single rate covers everything: the phone line and number, inbound and outbound calls, the voice the agent speaks with, speech recognition, the AI itself, recordings, transcripts, live transfer to your team, and the dashboard. There are no separate component charges underneath and no monthly platform fee. If you bring your own telecom account instead of using the number we provision, the 10¢ is unchanged and your carrier bills you directly for their per-leg cost.

What is included in the per-minute price?

Everything needed to run the call. On the default setup that means the phone number and line, calls in both directions, text-to-speech, speech recognition, the language model, call recordings, written transcripts, live warm transfer to a person, and dashboard access. Competitors commonly quote a platform fee and bill speech and model usage separately; there is no equivalent second bill here. The one exception is bring-your-own telecom: if you supply your own carrier account, that carrier invoices you for their per-leg cost and everything else stays inside the rate.

Are there contracts or minimum commitments?

No contract and no monthly commitment. There is exactly one floor, and it is on the payment rather than on usage: the smallest top-up the platform accepts is $20, because card processing fees consume too much of anything smaller. Beyond that Rexa.ai is pay-as-you-go — you hold credit and spend it at 10¢ per minute, you can stop at any time, and unused credit stays available until it expires. At enterprise volumes, discounts become substantial and are worth discussing directly with sales.

How do prepay tiers and bonus minutes work?

Top-ups start at $20, and prepays of $100 or more earn bonus credit on top — $100 is the first bonus tier, not an entry fee. The bonus lowers your effective rate: from +10% on a $100 prepay, giving an effective 9.1¢ per minute, up to +300% at $50,000, giving an effective 2.5¢. You are still charged 10¢ per minute; you simply have more credit to spend. A single checkout session takes up to $5,000, so larger balances are funded across several sessions or arranged with sales.

Do credits expire?

Credits expire one year from the date of purchase. The clock starts on the prepay date, and any bonus minutes earned on that prepay inherit the same expiry. Topping up does not extend the expiry of credit you bought earlier — each prepay carries its own one-year window.

Is there a free trial?

Trying Rexa in the browser is free and needs no card — a short test room, a voice, a prompt, and you are talking to the agent. Phone calls are the part that costs money: there are no free credits and no trial minutes, so placing real calls requires funding the account first. The smallest top-up is $20, which at 10¢ per minute is roughly 200 minutes of conversation — enough to test on real numbers before deciding anything. There is no contract, so that first top-up is the whole commitment.

How is call time measured?

By minutes of actual conversation. Time spent ringing an unanswered number is not conversation time. Browser-based conversations are charged at the same 10¢ per minute as phone calls — there is no separate price list for WebRTC.

How does Rexa.ai pricing compare to Vapi, Bolna and Deepgram?

The rates measure different scopes, so headline comparison misleads. As of 8 August 2026, Vapi publishes 5¢ per minute as a platform fee with model costs passed through at cost, Bolna publishes 6¢ with volume discounts, and Deepgram’s Voice Agent API runs 7.5¢ to 16.3¢ without telephony. Rexa’s 10¢ includes every component. Our cost breakdown page models this properly.

Getting started

How long does it take to set up a voice agent?

Most teams get a first agent placing real calls the same day. You pick a template or write a script, choose a voice, connect a phone number, and test. The work that takes longer is not setup but iteration: refining the script against real recordings usually takes a week or two of calls to settle.

Do I need developers to use Rexa.ai?

Not for a standard agent. Templates, scripts, voices and calling rules are configured in the dashboard. You need a developer when you want the agent to read or write data in your own systems mid-call, which uses the functions and webhooks APIs, or when you want to trigger calls programmatically from your own application.

What is an agent template?

A pre-built agent for a common job — recruiting screens, lead qualification, appointment reminders, support answering, surveys, collections, inbound reception — with the script, questions and outcome fields already defined. You edit it rather than starting from a blank page. Templates are a starting point; the script should end up specific to your business.

Can I test an agent before calling real people?

Yes. You can talk to any agent in the browser over WebRTC before it ever dials a phone number, and you can place test calls to your own numbers. Reviewing recordings and transcripts of test conversations is the fastest way to find the places a script goes wrong.

Phone numbers and telephony

Are phone numbers included?

Yes, by default. Rexa.ai provisions the phone side for you and the carrier cost is inside the 10¢ per minute — there is no separate charge for the number or the line. This is a real difference from platforms that quote a per-minute AI rate and leave you to buy and manage telephony separately. Bringing your own telecom account remains supported as an option, and on that path you keep your existing numbers and your carrier bills you for their per-leg cost.

Can I bring my own phone numbers or carrier?

Yes, though it is the option rather than the default — if you do nothing, we provision the number and the carrier cost sits inside the rate. If you already have a Telnyx, Twilio or Plivo account and want to keep your existing numbers, you can bring it and run calls on them. There is no additional platform charge for doing so, and no discount either — the per-minute rate is the same whichever way you connect. The difference is who pays the carrier: on bring-your-own, your carrier invoices you directly for their per-leg cost instead of us absorbing it.

Can the agent use my existing business number?

Yes, by bringing your own telecom account. Using a number your contacts already recognise generally improves answer rates compared with an unfamiliar one, so it is worth doing where you can. Inbound and outbound share the same number, so a contact who calls back reaches an agent rather than a dead line.

What happens if someone calls back the number that rang them?

They reach an inbound agent. Inbound and outbound run on the same platform at the same rate, so a returned call is answered rather than lost. This matters more than it sounds: a meaningful share of outbound campaign contacts call back rather than answering live, and an unanswered callback wastes the original contact.

Inbound and outbound calling

Does Rexa.ai handle both inbound and outbound calls?

Yes, on the same platform, with the same agents and the same 10¢ per-minute rate. Many platforms specialise in one direction and bolt on the other. Running both matters operationally because outbound campaigns generate inbound callbacks, and handling those on a separate system loses the context of the original call.

How many calls can run at the same time?

Concurrency is configurable per account and can scale well beyond what most teams need. The practical constraint is usually your own transfer capacity rather than the platform: generating more live transfers than your team can answer produces queued callers, which is a worse experience than the agent booking a callback.

What are retry rules?

Settings that control how many times an unanswered number is called again, how far apart the attempts are, and during which hours. Varying the time of day across attempts recovers more connections than simply attempting more often. Every attempt is logged whether or not it connected, within your permitted calling windows.

What is a structured call outcome?

A defined field recorded for every call — interested, not interested, wrong number, callback requested, transferred, no answer — plus any data captured in the conversation, such as availability or budget. Because outcomes are structured rather than free text, campaigns can be counted and lists re-targeted automatically instead of a human reading notes.

Can the agent answer calls outside business hours?

Yes, and out-of-hours coverage is often the clearest first use case because the alternative is nobody answering. The agent can resolve routine requests, book appointments, or capture the request with a callback preference for the next working day, so the call is not lost to voicemail.

Transfers and handoff

Can the AI transfer a call to a human?

Yes. Live warm transfer is included in the base rate. When the caller asks for a person, or the agent hits an intent you have marked as human-only, the live call is transferred to an available team member with the conversation context passed along, so the caller does not have to repeat themselves.

What happens if nobody is available to take the transfer?

The agent captures a callback preference rather than dumping the caller into a queue or voicemail, and the request is logged with the transcript attached. Whoever picks it up starts with the full context of what was already discussed instead of beginning cold.

Will callers always be able to reach a human?

That is your configuration choice, and we strongly recommend always honouring a request for a person. Callers accept an AI that resolves their problem quickly, and resent one that traps them. A reliable escalation path tends to raise satisfaction rather than lower it, because the calls reaching humans are the ones that needed one.

What is callback continuity?

When a conversation cannot be completed in one call, the context carries forward to the next one. The follow-up call — whether the agent places it or the contact calls back — starts from what was already established rather than restarting the script from the beginning.

Recordings and transcripts

Are calls recorded and transcribed?

Yes, and both are included in the per-minute rate rather than charged as extras. Every call produces an audio recording and a written transcript, retained against the call record alongside the structured outcome. Recording is subject to the consent rules configured for the jurisdiction you are calling.

How is recording consent handled?

Consent is captured at the start of the call, before the substantive conversation, and the consent state is recorded against the attempt. Where a jurisdiction requires all-party consent, the call can be configured not to record if consent is refused, while still proceeding as a conversation.

How long are recordings kept?

Retention is configurable to match your own data-retention policy and the obligations that apply to your industry. Because you are the data controller for call content, the right retention period is the one your policy specifies rather than a platform default. See the security page for storage and access controls.

Integrations and API

Does Rexa.ai have an API?

Yes. The API places and manages calls, retrieves recordings, transcripts and outcomes, and configures agents. Webhooks push call lifecycle, transcript and outcome events to your systems as they happen. Official Node.js and Python SDKs are available. See the documentation for the full reference.

Can the agent look things up in my systems during a call?

Yes, using functions. You expose an endpoint, describe it to the agent, and the agent calls it mid-conversation when it needs the data — checking an order status, reading availability, or writing a booking back. The caller experiences it as the agent simply knowing the answer.

Does it integrate with my CRM or ATS?

Call outcomes and transcripts are emitted as structured events over webhooks, so they can be written into any system that accepts an HTTP callback — which in practice covers every mainstream CRM and ATS, directly or through an automation layer. There is no requirement for a pre-built connector.

Can I trigger a call from my own application?

Yes. Calls can be created through the API, so a form submission, a CRM stage change, or any other event in your systems can place a call immediately. For lead response, this is the difference between calling in seconds and calling whenever somebody works the queue.

Latency and voice quality

How fast does the agent respond?

Fast enough that the conversation does not feel broken — response latency needs to sit within the length of a natural conversational pause, or callers start talking over the agent. We do not publish a single latency figure, because a number measured on our network under our conditions would not predict yours. Test it on your own calls.

Why does latency matter in a voice conversation?

Because there is no visual cue that the other party is composing a reply. In text, a delay is invisible; on a call, silence past about a second reads as the line dropping or the agent not understanding, and the caller speaks again. That produces overlapping speech and a conversation that feels wrong regardless of answer quality.

What happens on a poor-quality line?

Recognition accuracy degrades with line noise and strong accents, exactly as it does for human listeners. The right mitigation is a clean escalation path: configure the agent to transfer to a person after repeated misunderstanding rather than continuing to guess. An agent that gives up gracefully is better than one that insists.

Can I choose the voice?

Yes. A library of voices is available across accents and speaking styles, and the choice is included in the per-minute rate rather than priced by voice tier. The voice library page lets you hear each one before assigning it to an agent.

Languages

What languages does Rexa.ai support?

Rexa.ai is English-first. If your primary use case is calling in another language, test it thoroughly on your own accents and call conditions before committing volume — and be aware that some competitors specialise regionally. Bolna, for example, publishes support for more than ten vernacular Indian languages, which is a genuine advantage for that market.

Can one agent handle more than one language on a call?

Language handling is configured per agent. Where callers may switch language mid-conversation, test that path explicitly rather than assuming it, and configure escalation to a human for the cases it does not handle well. Code-switching is one of the harder things for any voice system to do reliably.

Security

How is call data secured?

Calls travel over encrypted connections and recordings, transcripts and call data are stored with access controls and encryption at rest. The security page documents the controls in detail, along with our responsible-disclosure policy and the contact path for reporting a vulnerability.

Can Rexa.ai be self-hosted or run in our own cloud?

Self-hosted and VPC deployments are available for high-volume customers, with terms agreed directly. If deployment location is a hard requirement from your security review, raise it early — it affects commercial terms and is not a self-serve option.

Who owns the call data?

You do. You are the data controller for the content of your calls and for the contact data you upload; Rexa.ai processes it to deliver the service. That distinction matters for your own privacy notices and retention policy, both of which need to reflect that calls are recorded and transcribed.

How do I report a security vulnerability?

Through the responsible-disclosure process on the security page. Reports are acknowledged within 48 hours, and good-faith research is covered by safe-harbor terms. Please report through that channel rather than publicly, and we will keep you updated through to the fix.

Compliance

Is AI phone calling legal?

It depends on your jurisdiction, your contact list, and the consent you hold — not on whether the caller is automated. Calling windows, do-not-call checks and recording consent are enforced by the platform, and every attempt is evidenced. You remain the caller of record and are responsible for confirming your campaign is permitted. This is not legal advice.

What compliance controls are built in?

Calling-window enforcement by jurisdiction, do-not-call checks before dialling, recording-consent capture at the start of the call, voice-clone consent handling, and a retained audit evidence trail of every attempt and outcome. These are enforced by the platform rather than by the script, because a control that depends on a script being correct is not a control.

Do I have to tell people they are speaking to an AI?

In several jurisdictions, yes — and we recommend disclosing everywhere regardless of whether it is required. Disclosure is configured in the opening line of the script. Callers told up front generally accept it; callers who work it out mid-conversation react considerably worse, and that reaction tends to overshadow the purpose of the call.

What is the audit evidence trail?

A retained record for every call attempt: the time, the outcome, the recording and transcript, and the consent state at that moment. It exists so that when a specific attempt is challenged, you can produce what actually happened rather than reconstructing it from separate systems after the fact.

How is voice cloning consent handled?

Every voice clone requires a scripted consent capture from the speaker, is held for human review before use, and can be revoked by the speaker at any time. Revocation deletes the clone and all derived assets within 24 hours.

Does Rexa.ai handle GDPR obligations for me?

No — it provides the controls you need to meet them. You are the data controller for call content and contact data, so your privacy notice, lawful basis, retention period and data-subject request process remain yours. The compliance page sets out the split of responsibilities in detail.

Still need an answer?

These answers were last reviewed on . If something here is out of date or unclear, tell us on the contact page and we will correct it.