Agent

A voice AI agent
inside your app.

Extentos Agent drops a voice AI agent into your mobile app, with smart glasses as the interface.

Your users talk; the agent answers — using your app’s own data and driving the glasses. You write the agent; Extentos runs everything around it: turn-taking, interruptions, speech in and speech out.

  • Extentos runs the conversation.

    Your model does the talking. Extentos is the runtime around it — streaming audio to and from the glasses and keeping the session alive for hours, past the provider's time limit. You just write tools.

    See the roadmap
  • Your model. Your voice.

    Any LLM through the Extentos AI Gateway or your own key — or a local model, which is free because it runs on the user's phone. Voices swap the same way, and there are free local voices too.

    Local models
  • Calls your code, drives glasses.

    Expose your own functions and the glasses' camera, video and audio as tools. The agent calls only what you wire up, and answers questions specific to your product.

    Capability vocabulary
Local models

Or run the model on the phone, free.

A cloud voice model charges you for every second it listens, and an assistant on your face listens constantly — so your bill grows with the engagement you were trying to win. A local model is free instead. Extentos runs the whole conversation on the user’s phone: speech in, the model thinking, speech out, with no server to pay for. Same handler code, same tools, same events. Only the brain moves.

  • Published for iOS and Android — one dependency line, not a research project
  • Automatic serves the best model each phone can genuinely sustain, and the cloud when none fits
  • Eleven neural voices that run on the phone too — and it all keeps working with no signal

Extentos Agent is live today.

The assistant runtime already ships in the Extentos SDK — connect the glasses, pick your model and voice, capture audio and photos, and ship to production, today.

agent prompt

38 tools·Node 20+Browse the tools