One principle, several stacks
Whatever framework you build on, the integration follows the same shape: get a copy of the caller’s audio to the verification service, and make the result available to the agent before it acts. What differs is where each framework exposes the audio and how the agent calls out. Integrations are being built with design partners; REST and WebSocket work with any platform that can reach the call audio.
Vapi
Vapi assistants can call custom tools. Register a verification tool and instruct the assistant to call it before sensitive steps — the tool returns the latest trust result for the caller. For continuous scoring, stream the call’s live audio to Vosco so a result is already waiting when the tool is called.
LiveKit
In LiveKit, the caller is a participant publishing an audio track — whether they joined over WebRTC or through SIP telephony. Your LiveKit agent can subscribe to that track and forward frames to the Vosco API, reading trust signals next to speech-to-text in the same session.
Pipecat
Pipecat, the open-source voice agent framework, models a bot as a pipeline of processors that pass frames along. A verification processor fits naturally: it consumes audio frames, sends them to Vosco and emits trust results downstream that your bot logic can branch on.
Meetings with Recall.ai
Voice-clone attacks are not limited to phone calls; impersonating an executive on a video call is a well-known fraud pattern. Recall.ai provides meeting bots for Zoom, Google Meet and Microsoft Teams with real-time participant audio. Streaming that audio to Vosco lets you verify who is speaking in the meeting.
Any agent, via MCP
The Vosco MCP server exposes verification as a tool to Claude, Cursor and any MCP-compatible agent, so the agent can check the voice mid-conversation whenever the stakes rise. For the telephony side, see adding verification to a Twilio voice agent.
Comparison
| Stack | Where the audio is | How the agent uses the result |
|---|---|---|
| Vapi | Live call audio | Tool call before sensitive steps |
| Twilio | Media Streams | Your service checks before acting |
| LiveKit | Caller’s audio track | Agent reads signals in-session |
| Pipecat | Audio frames in the pipeline | Trust frames downstream |
| Recall.ai | Real-time meeting audio | Flag or verify speakers |
| MCP | Provided by the host | Tool call |
All of them return the same signals. See the integrations overview or the developer docs.
Frequently asked questions
Can I add caller verification to a Vapi assistant?
Yes. Register verification as a tool the assistant calls before sensitive steps, and stream live call audio to Vosco so a result is ready when needed.
How does Vosco work with LiveKit Agents?
Your LiveKit agent subscribes to the caller’s audio track and forwards it to the Vosco API, then reads the trust signals in the same session.
Can Vosco verify speakers in Zoom or Teams meetings?
Yes, through Recall.ai meeting bots, which provide real-time participant audio from Zoom, Google Meet and Microsoft Teams.
Verify the voice before your AI acts.
Start on the Vosco API with 1,000 minutes free every month, or talk to us about on-device deployment.