Rufen is a multi-tenant voice AI platform built from Leipzig. BYOK from day one. EU-default.
Every voice AI platform I tried in 2025 had the same three problems: opaque pricing with model markup baked into every minute, prompts that lived in black-box dashboards with no audit trail, and a hard ceiling on which models you could actually run.
I wanted the opposite. Bring your own keys. Pay the providers directly. Version every prompt. Trace every call. Pick the LLM, STT, and TTS that work best for your accents and your budget — or run the local stack on Apple Silicon for sub-second latency without a single cloud round-trip.
So I started building Rufen — solo, from Leipzig, on nights and weekends until it became the whole job. We're in private beta with the first design partners. If that sounds like the platform you've been looking for, the waitlist is open.
Our mission
Rufen is a one-person operation today. Every line of code, every prompt, every reply — it's me.
I build the whole platform — Django backend, React console, Apple-Silicon voice worker, Asterisk dialplan, marketing site, legal pages. Background: AI engineering and data science. 2 months of solo work.
BYOK. Pay providers directly. We never markup AI.
Every prompt versioned. Every call traceable. No black boxes.
Your phone numbers, your customer relationships, your data.
Sub-1s latency. 22 voices. 13 languages. Sounds human.
Rufen is engineered in Leipzig — Saxony's quietest tech hub and a 70-minute train ride from Berlin. Close enough to the German enterprise market to ship at GDPR-grade by default, far enough from the SF bubble to focus on customers who actually pick up the phone.
Have a question about the product, the company, or how we handle your data? Reach out — most replies inside one business day.