From audio to a trust decision in minutes.
The Vosco API scores live audio as it streams. Send a call, an agent turn or a meeting over REST, WebSocket or MCP — get identity, liveness, replay and synthetic signals back in real time.
Four steps.
One integration.
- Get an API key in the self-service console — 1,000 minutes free every month.
- Send audio — a clip over REST, a live call over WebSocket, from your app with the iOS or Android SDK, or let your agent call Vosco through MCP.
- Read the signals — speaker identity, liveness, replay, synthetic and a combined trust score.
- Gate the action — let everyday turns through, step up before a payment, transfer or reset.
# Score a recorded clip
curl https://api.vosco.ai/api/v1/audio/analyze \
-H "Authorization: Bearer $API_KEY" \
-F "file=@caller.wav"
# → job_id; poll /api/v1/audio/analyze/jobs/{job_id}// Live stream: score while the conversation runs
const api = "api.vosco.ai/api/v1/audio";
const headers = { Authorization: "Bearer " + KEY };
const res = await fetch("https://" + api + "/sessions",
{ method: "POST", headers });
const { session_id } = await res.json();
const url = "wss://" + api + "/stream/" + session_id;
const ws = new WebSocket(url, { headers });
mic.on("audio", pcm => ws.send(frame(pcm)));
ws.onmessage = ({ data }) => {
const verdict = JSON.parse(data).summary.liveness;
if (verdict.includes("live")) agent.proceed();
else agent.stepUp(); // ask for more before acting
};{
"summary": {
"liveness": ["live"],
"synthesis": 0.97,
"replay": 0.96,
"snr_db": 21.5
},
"segments": [
{ "start_s": 0, "end_s": 5,
"labels": ["speech", "live"],
"age": 34, "gender": "male" }
]
}Three ways in.
Same models, same signals, same response — pick the interface that fits your pipeline.
REST
AvailableSend a recorded clip and get a verdict back. Good for voicemails, onboarding recordings and batch checks.
POST /api/v1/audio/analyzeWebSocket
AvailableStream live call audio and receive trust signals continuously while the conversation runs.
wss://…/api/v1/audio/stream/{id}MCP
AvailableExpose verification as a tool to Claude, Cursor and any MCP-compatible agent — it checks the voice when the stakes rise.
Vosco MCP serverNative SDKs
for iOS and Android.
Verify the voice inside your own app. The SDKs capture microphone audio, stream it to the Vosco API and hand back the same trust signals — no audio plumbing to write.
iOS SDK
- Microphone capture and streaming built in
- Speaker, liveness, replay and synthetic signals in real time
- Enroll and verify a user’s voice in-app
Android SDK
- Microphone capture and streaming built in
- Speaker, liveness, replay and synthetic signals in real time
- Enroll and verify a user’s voice in-app
Works with your
voice platform.
Vosco listens to the same audio as your agent. Integrations are being built with design partners — REST and WebSocket work with any platform that can reach the call audio.
Vapi
Voice agentsRegister Vosco as a tool your assistant calls before sensitive steps, and score live call audio throughout.
Twilio
TelephonyFork the caller’s leg to Vosco with Media Streams — scoring starts in the first seconds of the call.
LiveKit
Real-time audioSubscribe to the caller’s audio track from your LiveKit agent, over WebRTC or SIP.
Pipecat
Agent frameworkAdd Vosco to the pipeline as a processor that emits trust results your bot can branch on.
Recall.ai
MeetingsVerify who is speaking on Zoom, Google Meet and Microsoft Teams from real-time participant audio.
Retell
Voice agentsRegister Vosco as Retell’s post-call webhook: every call recording is checked for a live, replayed or synthetic caller.
Beyond the cloud API.
The same models also run inside your own infrastructure and directly on the device — for teams that can’t send audio out, and machines that have to decide offline. Both are set up with our team.
Vosco Precise inside your perimeter.
For banks, telcos and regulated teams that can’t send audio out.
Vosco SDK on the chip.
For robots and devices — runs offline with the 7M-parameter Vosco Nano model.
Start streaming in minutes.
1,000 minutes free every month, then $0.0042 per minute, lower at volume.