Give your AI agent a phone number.
The same agent that answers on your website can answer the phone — same corpus, same tools, same personality. Animam is the brain (an OpenAI-compatible Custom LLM). You bring the telephony: managed Vapi, your own SIP trunk, or any voice platform.
The brain, not the media stack
The hard part of voice isn’t SIP signalling — it’s the real-time media pipeline: speech-to-text, text-to-speech, turn-taking, barge-in, sub-second latency, fallback. Animam stays deliberately on the reasoning side of that line.
Animam owns
The reasoning: your corpus, your tools (booking, forms, escalation), personality, and a clean sentence-buffered SSE stream for the TTS.
You own
The media layer: SIP/carrier, STT, TTS, the latency budget, the numbers and the fallback — in the voice platform you choose.
Three ways to connect telephony
We provision it (no-code)
One POST and Animam creates the Vapi assistant and buys a phone number with the platform key. Your tenant never touches Vapi. Fastest to a working line.
You run your own Vapi & carrier
Import your own SIP trunk or numbers into your Vapi (Twilio, Telnyx, your own SIP), point the assistant at our Custom LLM URL, and link it. Your SIP lives in your Vapi — there is nothing SIP to configure on our side.
Any OpenAI-compatible voice stack
The endpoint is plain OpenAI-format SSE — not Vapi-specific. Vapi, LiveKit Agents, Pipecat, Retell, Bland or your own stack: point the model URL at Animam and own the SIP, STT/TTS and latency budget.
Full setup — the /voice/setuproutes, the Custom LLM contract and payload examples — is in the voice documentation.
Same agent, voice-adapted
- One knowledge base for chat and voice — update once, both reflect it.
- Tools run on calls: book a meeting, check availability, capture a form, escalate to a human.
- The agent reads dates, prices and addresses in natural speech, not formatted text.
Frequently asked questions
Does Animam handle the SIP / telephony itself?
No — by design. Animam is the brain: an OpenAI-compatible Custom LLM endpoint that reasons over your corpus and runs your tools. The telephony layer (SIP, STT, TTS, turn-taking, latency, fallback) is owned by your voice orchestrator. That separation is what lets you bring your own carrier without us standing between you and your numbers.
So how do I use my own SIP trunk?
Import it into your voice orchestrator (Vapi supports BYO carriers — Twilio, Telnyx, raw SIP), point that assistant’s Custom LLM URL at https://api.animam.ai/voice/your-slug, and link it to your tenant with PUT /voice/setup. Your SIP never touches Animam; we just answer the reasoning.
Which voice platforms work?
Any that accept a Custom LLM URL in OpenAI format: Vapi, LiveKit Agents, Pipecat, Retell, Bland, or your own stack. Animam is provider-agnostic on purpose — the same endpoint serves all of them.
Is it the same agent as my website chat?
Yes. Same corpus, same segments, same system prompt, same tools — BOOK_MEETING, CHECK_AVAILABILITY, SUBMIT_FORM, ESCALATE_TO_HUMAN all work on a call. The agent adapts its phrasing for speech (reads dates, prices and addresses naturally). Update your knowledge base once; both channels reflect it.
What about latency and audio quality?
That lives in your voice layer, which is the right place for it. Animam streams responses sentence-by-sentence (SSE) so the TTS engine gets clean phrases instead of token fragments. The end-to-end latency budget — STT, network, TTS — is tuned in your orchestrator.
Is voice data GDPR-friendly?
Animam is hosted in France and only ever sees the transcript text, not the audio. Where the audio and recordings live depends on your voice orchestrator — another reason builders who care about data residency keep the media layer in their own account.
Why not a fully native SIP option from Animam?
Because terminating SIP and running the media stack is a telecom product in itself (SBC, RTP, codecs, fraud, E911) — and specialists like Vapi and LiveKit already do it well. Staying the brain keeps Animam focused and keeps you free to choose (or change) your voice layer.
Put your agent on the line
Hosted in France, GDPR-ready. Bring your voice layer, keep the brain in one place. Start free.