Voice agents act. Verify the human first.
Vosco checks the voice on the line before your agent accesses an account, makes a booking or moves money — who is speaking, and whether the voice is live, cloned or replayed. In real time, from the first seconds of the call.
The transcript can’t tell a
customer from a clone.
An AI voice agent answers, listens and acts. It has no way to separate a customer from a cloned voice, or a live human from another model on the far end of the line — the transcript reads identically either way.
Cloned callers
A clone built from seconds of public audio passes every knowledge check the agent makes.
Replayed audio
A recording of the right customer clears a voiceprint gate as easily as the customer does.
Machine-speed attacks
Synthetic callers probe agents thousands of times — no fatigue, no cost.
Audio reaches the model
The agent acts on what it heard, not on who said it.
Four questions,
answered in one call.
Speaker recognition, trust scoring and voice insights are ours end to end — trained in-house and benchmarked in public. Every call returns all of them at once.
- Voice vectorA biometric profile unique to each voice
- 1:1 verificationIs this the speaker they claim to be?
- 1:N identificationWhich enrolled speaker is this?
- LivenessLive human vs. recording or replay
- GenAIFlags synthetic and cloned speech
- Audio qualityStudio, phone or noisy channel
- GenderEstimated speaker gender
- AgeEstimated age range
- EmotionNeutral, happy, sad or angry — on request
- LanguageSpoken language
Plugs into the voice stack
you already run.
Vosco sits beside your telephony and your agent framework — it listens to the same audio and hands the agent a trust decision. No re-platforming, no extra hop in the conversation.
Register Vosco as a tool your assistant calls before sensitive steps, and set Vosco as the assistant’s custom transcriber to score the caller throughout the call.
Fork the caller’s leg to Vosco with Media Streams. Scoring starts in the first seconds and runs alongside the call — it never holds the call up.
Subscribe to the caller’s audio track from your LiveKit agent — WebRTC or SIP — and read trust signals next to speech-to-text in the same session.
Add Vosco to the pipeline as a processor that consumes audio frames and emits trust results your bot logic can branch on.
Stream real-time participant audio from Zoom, Google Meet and Microsoft Teams to verify who is speaking — and catch a cloned executive on a video call.
The Vosco MCP server exposes verification as a tool to Claude, Cursor and any MCP-compatible agent, so it can check the voice mid-conversation whenever the stakes rise.
Cloud, or inside
your perimeter.
Agent platforms, contact centres and fintech stream audio in and get identity, synthetic-speech and replay signals back in milliseconds — priced per verified minute.
- Start on the cloud API — REST, WebSocket or MCP, with 1,000 minutes free every month.
- Move on-prem when you need to. Banks, telcos and regulated buyers run Vosco Precise inside their own perimeter on an annual licence.
- Keep one policy model. The same signals and response shape run in the cloud, on-prem and on the device.
Vosco Precise
| Speaker accuracy | 99.3% |
| Synthesis error | 3.1% |
| Replay error | 2.0% |
| Latency (GPU) | < 10 ms |
| Throughput | 1,000+ streams |
| Interfaces | REST · WebSocket · MCP |
| Pricing | Per verified minute |
How it works on the call.
Vosco listens alongside your speech-to-text for the whole call and returns a fresh verdict every 3 sec — inference under 10 ms, so it keeps confirming it’s the same caller with no pause in the conversation.
- Stream
Analyse the caller’s audio
Send the call audio to Vosco alongside your speech-to-text — over WebSocket, or through your platform’s integration.
- Verify
Trust signals, mid-call
Identity, liveness, replay and synthetic signals land while the agent is still on its first turn — and keep updating.
- Act
Gate the action, not the call
Let ordinary turns through; require a passing trust score before a transfer, a payment or a credential reset.
Verify the caller before the agent acts.
Start on the free tier — 1,000 minutes a month — or tell us about your platform and we’ll map the integration with you.