GStudio · AudioAgent

Build AI voice agents that talk like people

Low latency and easy to interrupt, it looks things up, calls tools and sees the screen while it talks. Serve customers on the web, in GWork and by phone, with every call recorded for playback.

Why AudioAgent

  • As natural as a person

    Fast replies, and callers can cut in at any time. It knows when someone has finished speaking, fills awkward silences, and speaks 40+ languages.

  • Gets it done on the call

    It searches your knowledge base, calls API, MCP and OAuth tools and queries data tables mid-call, so order checks, bookings and sign-ups are handled in a single call.

  • Fits your channels and team

    Serve customers on the web, in GWork and by phone: connect your Twilio number to answer calls. Outbound calling and handoff to a person with full context are coming soon.

Three voice engines for different scenarios

Switch engines on the same AudioAgent at any time; knowledge, tools, data tables, the welcome guide and call control work with all three, and all three can see the camera or a shared screen.

  • Real-time model

    Best for: Support and advice where low latency and natural conversation matter most.

    One real-time model listens and speaks directly and decides whose turn it is, for the lowest latency and the most natural conversation, with a choice of output voice.

  • ASR-LLM-TTS

    Best for: Voice support with complex business rules that needs a specific text model.

    Speech is transcribed, handed to a text model of your choice and read back with speech synthesis. The knowledge base is searched every turn, for the most control and the highest latency.

  • LLM+TTS

    Best for: Balancing understanding with synthesized voice quality.

    A model that understands audio takes speech directly, and speech synthesis reads the reply; knowledge, tools and databases are offered as function tools.

Everything a great call needs, ready in AudioAgent

Listening, speaking, looking things up and seeing, plus handoff, phone lines and post-call analysis, all configured in the console.

It tells when the user has finished by pauses or by meaning, and stops to listen the moment they cut in. It follows up on silence and hangs up on timeout or long silences. Add brand names and terms to the transcription instruction, and the real-time engine understands whatever language the user speaks.

Enterprise-grade voice agents

Trust AudioAgent with your customers

Call data under control, clear permissions and transparent costs, with the option to deploy in your own environment.

  • 01

    Data security

    Every incoming call is signature-checked, integration credentials are stored encrypted, and web calls use short-lived tokens. Private deployment keeps data in your environment.

    • Call signature checks
    • Encrypted credentials
    • Private deployment
  • 02

    Access and audit

    SSO sign-in and role-based permissions, with every call’s record and runtime sequence available to review.

    • SSO
    • Role permissions
    • Call logs
  • 03

    Transparent usage

    Credits are charged by transcription time, synthesized characters and model tokens, while phone line charges go through your own Twilio account. Concurrency is capped by plan, and iframes can set credit limits.

    • Usage-based billing
    • Concurrency limits
    • Credit limits

AudioAgent at a glance

  • 3voice engines
  • 40+languages
  • 30+preset voices
  • 4channels

AudioAgent FAQ

AudioAgent is the real-time voice agent in GStudio. It talks with customers live on the web, in GWork and by phone, and can be interrupted at any time. During the call it searches your knowledge base, calls tools and queries data tables to resolve issues on the spot, and it can see the camera or a shared screen. Every call is recorded for playback.

Build Smarter AI Agents Right Now

Get 100 Free Credits Monthly!