Voice infrastructure for autonomous agents
Private beta. The SDK and KYA agent identity are live today. Live call access is granted by request.
VoiceGate is clean voice infrastructure for AI agents. Your agent picks the number, voice and language — we place the call, transcribe what the other side says, and speak your agent's replies in real time. WHAT to say is decided by your agent's brain, not us. Billing is per minute, in crypto.
We don't give the agent a "brain" — you already have that. We give it ears, a mouth and a phone line.
One API call: number, voice (male / female), language and your brain-agent's webhook. Top up the USDC balance before the call.
We place the call over our telephony. The other side's speech → we transcribe (STT) → send to your agent → speak its reply (TTS). Real time, ~1 s.
We charge the balance for actual minutes — USDC stablecoin on the Base network. No subscriptions, no cards, no people.
Hear it in action
Real example dialogues. Your agent decides what to say — VoiceGate speaks it in the caller's own language, inbound or outbound. These are shown in the language each call was made in.
↑ Same engine, any language and voice. WHAT to say comes from your agent; VoiceGate is the ears, mouth and phone line.
50+ languages
Your agent picks the STT + TTS language before each call — one field in the API. Real-time transcription and speech in 50+ languages, each with a natural male or female voice, powered by Google's speech engine. Ukrainian, Polish and Russian included.
Pick any of the above per call via the language field. More added as Google's voice coverage grows.
Machine-native API
Call API — private beta. The request shape below is the reference contract. The SDK and KYA identity flow work today; live call credentials are issued on beta access.
# the agent hires a voice POST /v1/agent-call x-api-key: vg_live_… { "to": "+1...", # destination number, E.164 "greeting": "Hi, this is Ada…", # spoken on connect "prompt": "Confirm the 3pm…", # drives the agent's side "language": "en", # en, es, fr, de, uk, pl, ru, ar, hi, zh … 50+ "voice": "female" # male | female } # ← while balance > 0 — the line is yours { "status": "calling", "call_uuid": "8f3c1e6a-…", "hold_micros": 600000, # reserved, not charged "rate_micros_per_min": 60000 # $0.06/min, billed per second }
New service
AI answers and makes calls for you: it handles your customers around the clock and calls your database — reminders, confirmations, callbacks. Voice and language of your choice, billed per minute.
Takes calls, answers FAQs, qualifies, books appointments, captures data and passes hot leads to you. Not a single missed call.
To YOUR contacts who expect you: appointment and payment reminders, confirmations, order status, callbacks, customer re-activation, short surveys.
Every call — recording, transcript and outcome (reached / agreed / call back). Integrates with your CRM. Pay for minutes talked, not headcount.
⚠ Only to your own contacts and with the recipient's consent. We don't do cold mass-calling or spam — see the policy.
🇺🇸 For the US — TCPA / DNC compliant: every list is scrubbed against the National Do Not Call Registry, marketing calls only with written consent.
Get started
Fill the brief — we'll come back with a plan, a voice sample and a price. No spam.
The carrier's rate for the destination plus a flat VoiceGate margin of $0.06 / min. No subscriptions, no hidden fees. Speech-to-text (STT) and text-to-speech (TTS) included.
Where to send payment
Top up the agent's balance in USDC on the Base network. Your dedicated top-up address is issued during onboarding, once beta access is granted — never send funds to an address posted publicly.
🔒 address issued at onboarding
⚠ USDC on the Base network only. We never publish a wallet address — anyone who does is not us.
Policy — read before integrating
We are a deliberately narrow service. Abuse burns SIMs, reputation and breaks the law — so the boundaries are strict from the first call.
FAQ
VoiceGate is voice infrastructure for AI agents: a phone line with built-in speech-to-text and text-to-speech. It gives your agent ears, a mouth and a real number to call from. Your agent's model decides what to say; VoiceGate carries the call.
One API call sets the number, voice, language and your agent's webhook. On every turn we transcribe what the other side says, POST that text to your webhook, and speak your agent's reply back in real time. The brain stays yours.
VoiceGate is model-agnostic — you bring your own agent brain instead of a bundled one. It bills per minute in USDC on the Base network with no subscription and no card, and it adds KYA: verifiable agent identity and reputation on the open ERC-8004 standard.
From $0.06 per minute on top of the carrier rate for the destination, prepaid in USDC on the Base network. Speech-to-text and text-to-speech are included. No subscriptions, no cards, no minimum commitment.
Calls run in multiple languages, including English, Spanish, French, German and Chinese, with a male or female voice chosen per call.
KYA is KYC for AI agents: a verifiable credential for identity plus a portable reputation score on the open ERC-8004 standard. VoiceGate only lets an agent onto the line once it has a verified identity and enough reputation.
Price = the carrier's rate for the destination + a VoiceGate margin of $0.06 / min. Speech-to-text (STT) and text-to-speech (TTS) included. Payment in USDC on the Base network.
| Destination | Landline / Mobile |
|---|---|
| USA | $0.07 |
| Canada | $0.07 / $0.12 |
| United Kingdom | $0.08 / $0.45 |
| Germany | $0.08 / $0.37 |
| France | $0.08 / $0.35 |
| Poland | $0.07 / $0.21 |
| Ukraine | $0.23 / $0.38 |
| India | $0.09 |
| Rest of the world | from $0.07 |
Outbound rates shown (agent calls out) — landline / mobile, since calling a mobile costs noticeably more. Incoming calls $0.07/min; your own number from $2/mo. Each rate = the destination carrier's rate + a flat $0.06/min, drawn per minute from the agent's balance. Full per-country table on request.