Introduction

Calleague is an AI voice-agents platform for real-time inbound and outbound phone calls, multilingual and CRM-integrated.

Calleague is an AI voice-agents platform. It runs real-time STT → LLM → TTS pipelines so software agents can handle real phone calls — both inbound and outbound — the way a trained human operator would: listening, understanding, answering, and acting on what the caller says.

Calleague is multilingual (32 languages, with Turkish as a first-class language), connects to the telephone network over SIP via the CalleKit voice stack, and integrates with your CRM and business tools so an agent can look things up and take action mid-call.

🌐

Learn more at calleague.ai and try a live voice agent in your browser at calleague.ai/try.

What Calleague does

A Calleague voice agent answers (or places) a phone call and holds a natural, low-latency conversation:

Speech-to-text (STT) transcribes the caller in real time.
A large language model (LLM) decides what to say and which tools to call.
Text-to-speech (TTS) speaks the response back to the caller.

This loop runs continuously for the duration of the call, with voice-hardening safeguards (call-duration limits, model timeouts, stall detection and re-prompting) tuned for production telephony rather than a chat demo.

Key capabilities

Real-time voice pipeline — streaming STT → LLM → TTS with latency budgeting for phone-grade conversations.
32 languages — multilingual recognition and synthesis; Turkish is first-class across the product and documentation.
SIP telephony — inbound and outbound calls over the public telephone network through the CalleKit SIP ↔ WebRTC bridge.
CRM integration & tool calling — agents can fetch records, look up data and trigger actions through connected business systems during a call.
Call recording — optional recording with two backends (agent-side WAV or composite egress), encrypted at rest, with consent handling.
Automation — a production-stable workflow module (schedule / manual / webhook / event / email triggers, multi-step pipelines, branching, notifications, governance and audit).
Flexible deployment — run Calleague as self-host, SaaS (multi-tenant) or on-prem.

Who Calleague is for

Use case	What a Calleague agent handles
Call centers	High-volume inbound calls, routing, FAQ deflection and after-hours coverage.
SDR / outbound sales	Outbound dialing, qualification and follow-up at scale.
Customer support	First-line resolution, ticket creation and escalation to humans.
Healthcare	Appointment reminders, scheduling and triage prompts.
Hospitality	Reservations, confirmations and guest inquiries.

How it is built

Calleague runs as a two-tier deployment — a Platform tier (UI, API, telephony bridge) and a GPU tier (the voice worker plus STT/LLM/TTS/embedding/rerank models) — sharing a single filesystem. Language-model traffic is brokered through Qevron, an OpenAI-compatible LLM gateway. Data lives in MongoDB (calleagueV2).

🛠️

Every service is installed, built, started, stopped, restarted and monitored through a single entry point: ./service.sh.

# bring services up and check that they are healthy
./service.sh start
./service.sh status
./service.sh health

Security & compliance

Calleague is designed for regulated, customer-facing telephony:

AES-256-GCM encryption at rest and TLS in transit.
Optional PII auto-redaction and configurable retention windows.
BYOK (bring-your-own-key) support.
Aligned with KVKK (Türkiye) and GDPR, including user consent and right-to-erasure flows, with full audit logs.

Next steps

Architecture — the two-tier topology, the full service map and how the pieces fit together.
Deployment models — choosing between self-host, SaaS and on-prem.
Quickstart — install and start Calleague for the first time.

Calleague is built by Arpanet Bilişim Teknolojileri A.Ş. — ODTÜ Bilişim İnovasyon Merkezi, Çankaya, Ankara, Türkiye. Contact: info@arpanet.com.tr · +90 312 486 00 22.