Is SIPHON really free?

Yes. SIPHON is 100% open-source under the Apache 2.0 license. You never pay us a platform fee. You only pay for your own infrastructure and the AI/Telephony providers you choose (e.g., OpenAI, Twilio).

Do I need to be a VoIP expert?

Not at all. SIPHON abstracts away the complex SIP signaling and media handling. If you're comfortable with Python, you can build production-grade voice agents.

Which AI models can I use with SIPHON?

You have total freedom. SIPHON supports OpenAI, Anthropic, Gemini, Groq, and open-source models via Ollama. You can swap providers with a single line of code.

How does SIPHON handle latency?

SIPHON is optimized for real-time performance. By hosting it on your own infrastructure (close to your users) and using WebSocket streams, you eliminate the additional hop and latency overhead of managed platforms.

Can I scale SIPHON to thousands of calls?

Yes. SIPHON is built on top of LiveKit, which powers massive scale streaming. The worker architecture allows you to horizontal scale your agent fleet to handle any volume.

How does SIPHON compare to closed platforms like Vapi or Retell?

Those are managed platforms that charge a markup per minute (often $0.05-$0.10/min). SIPHON gives you the same capabilities but runs on your own infrastructure, costing just the raw provider fees (typically ~$0.005-$0.02/min) and giving you full code control.

SIPHON

Back to Blog

architecture

1 post tagged with "architecture"

SIPHON: The Telephony-First Framework for Calling AI

frameworktelephonyarchitecture

January 28, 2026

SIPHON: The Telephony-First Framework for Calling AI

Building an AI that can chat via text is now a solved problem. Building an AI that can handle a real phone call in production is still a nightmare of fragmented infrastructure. SIPHON was built to end the plumbing.