Voice AI like Lego.
You build. We deliver.
REST API, SDKs, webhooks and bring-your-own models. Build voice AI into your platform, your workflow, your industry — on infrastructure that already works.
{
"name": "Sales Agent",
"model": "gpt-4o-mini",
"voice": {
"provider": "cartesia",
"voice_id": "de-empathic"
},
"language": "de",
"tools": [
"transfer_call",
"book_appointment"
]
}{ "id": "asst_8f2a…",
"status": "ready" }{ "duration": 124,
"transcript_url": "…",
"extracted": {
"email": "anna@…",
"appointment_at": "…"
}
}You're not building from zero.
You build on a layer.
Audio pipelines, telephony edge cases, LLM cost optimization, SIP trunk swaps — already done. You build your product, not our infrastructure.
Over 1,000,000 real calls, 7 model migrations, 3 SIP provider switches — all already done.
Realtime, full-duplex, WebRTC. Sub-100ms latency that makes your agent sound natural — not like an IVR.
Scales horizontally. Black Friday, election night, newsletter spike — we've seen it.
Multi-cloud redundancy with failover. Webhook retries with exponential backoff. Sentry tracing per call.
Three calls.
A fully working voice agent.
Create an assistant, start a campaign, receive webhook events — done. No audio pipeline, no telephony quirks.
// 1. Create assistant
POST /api/assistants
{
"name": "Sales Agent",
"system_prompt": "You are a helpful sales assistant for …",
"provider": "openai",
"model": "gpt-4o-mini",
"voice": { "provider": "cartesia", "voice_id": "en-empathic" },
"language": "en",
"tools": ["transfer_call", "book_appointment"]
}Battle-tested.
Self-hostable.
We self-host or you self-host — both possible. White-label via custom domains, branding and design systems.
LiveKit · Realtime
WebRTC + SIP bridging. Sub-100ms latency. Native telephony via Telnyx, BYO trunks or your own providers.
AI orchestration
STT (Deepgram, Whisper) → LLM (GPT-4o, Claude, Gemini, DeepSeek) → TTS (Cartesia, ElevenLabs). Modular, swappable.
Supabase · State
Postgres + Auth + Realtime. Multi-tenant, RLS, service roles for backend integrations.
AWS S3 · Media
Stereo MP3 recordings, pre-signed URLs, your own bucket possible. Full transcripts with speaker diarization.
Everything you expect.
Nothing you wait for.
REST API
Assistants, Calls, Campaigns, Contacts, PhoneNumbers, Recordings — all as clean resources.
Webhooks
call.started, call.completed, call.transferred, analysis.ready, lead.qualified — with 3 retries and exponential backoff.
Function tools
Define any tools (HTTP endpoints) your AI can call during the conversation.
Multi-provider AI
OpenAI, Anthropic Claude, Google Gemini, DeepSeek, Qwen, Grok, OpenRouter — freely selectable per assistant.
Multi-tenancy
One account per end customer, own keys, own limits, own branding. White-label ready.
Cron endpoints
Outbound engine, phone-number verification, trial expiration — cleanly documented and observability-ready.
From MVP
to multi-tenant platform
SaaS platforms
Ship voice AI as a feature to your customers. Own domain, own branding, own limits per account.
Agencies
Build voice bots for clients — without learning telephony infrastructure.
Industry specialists
Healthcare, legal, real estate — bring your own logic, your own tools, your own compliance.
Internal tools
Voice bots for internal workflows — onboarding calls, IT ticket creation, HR FAQ.
Once you've built with it,
you don't go back
"We built a voice AI layer for our insurance platform in 4 weeks — with the Talkturo API. Build vs. buy wasn't a discussion anymore."
"We deliver voice AI as white-label to 40+ mid-market customers — all via API. Multi-tenancy and SSO from day 1, no maintenance overhead."
"Sub-100ms latency was non-negotiable for us. Talkturo was the only provider that delivered — and with GDPR hosting in Frankfurt."
Free tier available · Enterprise SLA on request
Production-ready
from the first API call
We've spent 3 years learning telephony edge cases, LLM migrations and compliance — you get the result as an API.
Auth & RBAC
API keys per account, scoped tokens per team, service roles for backend integrations.
EU hosting & BYOC
EU data residency by default. Bring-your-own-cloud on request — fully self-hosted is also possible.
BYO SIP trunk
Telnyx, Twilio, your own trunk — we route. Phone numbers in 50+ countries.
SOC 2 Type II
Audit logs, access reviews, incident response. Compliant for regulated industries.
Webhooks with retries
3 retries with exponential backoff. Signed payloads. Idempotency keys on all mutations.
Sentry observability
Every call has a traceID. Every error is traceable — down to LLM roundtrip latency.
In 10 minutes
to your first live call
Get the API key, copy the quickstart snippet, deploy it. The first 100 minutes are on us.
- API key in under 1 minute
- Quickstart in 5 languages (cURL, JS, Python, Go, Ruby)
- Free tier with 100 min/month
- Solutions engineering on request
curl -X POST \
https://api.talkturo.com/v1/calls \
-H "Authorization: Bearer $TT_KEY" \
-d '{
"assistant_id": "asst_8f2a…",
"to": "+49123456789"
}'- 100 min/month free
- EU hosting standard
- Webhooks + retries
- Self-host possible