Voice-first agentic platform

Build Arabic voice agents that actually ship

Compose ALLaM reasoning, VoxCPM speech, and Telegram delivery into agentic workflows — without gluing together five SDKs.

Telegram SLA
<5s
First-class bilingual
EN · AR
Reasoning engine
ALLaM
Native Arabic voice
VoxCPM

Everything you need to run a voice agent

From the first spoken word to the final Telegram reply, the loop is wired end-to-end.

Arabic-native voice
VoxCPM streaming TTS and STT tuned for MSA and dialect, with seamless code-switching to English.
ALLaM reasoning
Plug ALLaM into any node in your graph. Use it for intent, extraction, planning, or final response.
Agentic flow runner
Topologically scheduled nodes with typed edges. Build a flow once, run it from chat, voice, or cron.
Telegram-ready
A managed webhook that meets the 5-second SLA and forwards voice notes straight into your graph.
Observability
Every run captured in Firestore — inputs, outputs, latency per node. Replay and audit in one click.
EN / AR with RTL
First-class bilingual UI. Switch languages without losing context — RTL layouts handled automatically.

Built on

  • Next.js 16
  • React 19
  • ALLaM
  • VoxCPM
  • Firebase
  • Telegram
  • shadcn/ui

Build Arabic voice agents that actually ship

Compose ALLaM reasoning, VoxCPM speech, and Telegram delivery into agentic workflows — without gluing together five SDKs.