Build voice agents on one API
Create an agent, attach your own carrier number, and it is answering and placing calls. Real-time speech, webhooks, function calling and Live Monitor, over a plain REST API. Flat $0.08 a minute, no credits, no per-feature upcharges.
Everything the phone layer needs, in one API
Building voice yourself means stitching speech recognition, a language model, a voice, and telephony into one low-latency loop, then keeping it alive under real traffic. Nixflex runs that whole pipeline on its own engine and hands you a clean REST interface over the top.
Answer inbound and place outbound calls, single or batched, from the same agent. Cold and warm transfers, mid and post-call SMS, voicemail handling.
Transcript, summary and sentiment on every call, by webhook or GET /v1/calls.
Custom functions let the agent hit your API and use the result to keep talking.
Listen to any live call in real time with Live Monitor, something only possible because Nixflex runs its own engine rather than forwarding audio to third parties.
Pricing you can predict before you ship
Most voice platforms quote a low headline number, then bill in credits that hide the real per-minute cost until you are already integrated. Nixflex is one flat rate: $0.08 per minute of voice, premium model included, your own carrier underneath. What you calculate on day one is what you pay at scale.
Built for products with many customers
If you are putting voice inside a SaaS or reselling to clients, the hard part is not one call, it is provisioning and isolating hundreds of them. On Nixflex an agent is a template that attaches to unlimited numbers, so you give each customer their own number backed by a shared agent, and manage the lot from one API key. The multi-tenant control plane other platforms make you build is the default here.
Explore the developer docs
Full endpoint reference lives in the docs. Coming from another platform? Nixflex is a modern alternative to Retell, Vapi and Bland. Compare them →
Developer FAQ
What do I need to start building?
An API key and one of your own carrier numbers. Create an agent with a POST request, attach a Twilio or Telnyx number to it, and the agent is answering calls. There is no SDK to install first; the API is plain REST, so you can call it from any language with an HTTP client.
Is there an official SDK?
Not yet. Nixflex is a REST API today, which means you can use it from any language with the HTTP tools you already have, and typed client libraries are on the roadmap. We would rather ship an honest REST surface than a thin wrapper, so the endpoints and payloads you see in the docs are exactly what you call.
How does pricing work?
A flat $0.08 per minute of voice, pay-as-you-go, with the premium model included. Numbers using Live Monitor bill inbound minutes at $0.09. There are no credits to buy, no per-feature upcharges, and no monthly platform fee to start. You bring your own carrier, so number and call charges stay on your own account where you can see them.
Can I manage many customers from one account?
Yes. An agent is a template that attaches to unlimited phone numbers, so you can give every one of your own customers their own number backed by a shared agent, all from one API key. That is how a single dental agent answers across two hundred separate clinic numbers today.
How do I get call data back into my system?
Register a webhook and Nixflex posts events as they happen, ending with the full transcript, a summary and a sentiment score when the call finishes. You can also pull any call from GET /v1/calls, and the agent can call your own APIs mid-conversation through custom functions.
Can I try it before wiring it into production?
Yes. Build an agent in the dashboard and call it yourself, or run a web call, to hear exactly what a caller would before a single customer reaches it. Because every change is prompt and settings, iterating takes minutes, not a redeploy.
What happens if a provider has a problem mid-call?
The platform fails over automatically. Speech recognition switches provider if one goes quiet, the language model falls back before the first word is spoken, and the voice has a fallback engine. You do not wire any of this yourself; it is part of running on the platform.
Ship your first agent today
Grab a key, create an agent, attach a number. Minutes, not months.