{{ t.body }}
Pick a number, pick a voice, choose the languages you serve, and point the line at the processes it can start. Everything a voice agent product gives you — sitting on Glove Core instead of its own private memory.
One thread in Tandem: WhatsApp, the call, the email that follows. Filter by channel, open the call and follow the turns — the router decision, the barge-in, the fields it wrote into state.
Nothing in that thread is a separate voice product. The call ran the same process the chat started, wrote into the same fields, and left the same audit trail. That only works if the voice pipeline and Glove Core are one system.
Most voice agents are a separate product with their own memory. Glove runs voice on its own pipeline and on the same shared state as chat and email — a call continues a WhatsApp conversation without losing a thing.
The same invisible handoff, now on calls. A human steps into a live call, gets guided in real time, and steps out. The caller never knows.
Q4 2026No. It runs on its own low-latency pipeline, but on the same Glove Core as chat and email. The call is another transport into the same process, not a separate product with its own memory.
Yes. The call opens with the process exactly where the chat left it, so the customer does not repeat an order number or explain the situation again.
The pipeline handles barge-in and returns to the step it was on. Because the flow is held by Glove Core and not by the model, an interruption does not send the conversation off-script.
Yes, with the whole state loaded — the same handoff as any other channel. Every transition, including the handoff itself, is logged.