Voice Agent
Hebrew-first AI voice agent with scenarios.
WebRTC voice agent powered by Gemini Live. Tool calling, scenario builder, mid-call resilience.
Generic voicebots speak English and miss context. Yours sound robotic in Hebrew — or worse, drop the call.
Most realtime voice platforms were built for English first and Hebrew last. They struggle with niqqud-free transcription, RTL prompts, and Israeli phone-system conventions. We built the opposite stack.
How it works
- 01Define a scenario
Visual scenario builder — prompts, tools, business hours, transfer rules. No code.
- 02Pick a provider
Gemini Live, OpenAI Realtime, or both with same-provider session resumption. Switch without code changes.
- 03Plug into PSTN or app
Inbound Twilio number, embedded WebRTC, or our mobile SDK. Sub-second latency end-to-end.
Capabilities
Every published scenario is immutable. Live calls pin a version — edits never affect running calls.
Your tools, voice-native. The agent can fetch CRM data, book appointments, send SMS — mid-call.
Network glitch? Session resumes on the same provider, same context, same tool state.
Native Hebrew voices, niqqud-aware prompts, mixed-language tolerance (Hebrew + English jargon).
Per-org provider config. Switch between providers per scenario, per model, per capability.
Searchable transcripts, sentiment trend, tool-call audit. All retained per your policy.
Use cases
Replace the "press 1 for English" tree with a real conversation that books, routes and confirms.
Call lists, dynamic scripts, real-time outcome capture — without burnout.
Inbound leads from the website get an instant voice callback that qualifies and books a human.
Frequently asked
Ready to see it run?
We deploy a pilot in a day. Hebrew-native, EHR-aware, ROI-measurable.
Book a Demo