White-label voice AI

Voice inside your product, under your name

For software companies that want to sell voice AI as part of what they already sell. Your brand, your dashboard, your pricing to your customers. Our orchestration layer underneath, and our engineers next to yours while you wire it in.

your customers never see us · per-customer usage and billing · EU-hosted

Three shapes

Pick how deep the voice goes

Embedding voice does not have to mean one thing. Most partners start with one of these and grow into the next.

web call

A talk button in your app

A browser or in-app voice conversation over WebRTC, with no phone number involved. Your backend asks our API for a session token and hands it to the front end, so the browser never touches an API key. Same pipeline and same billing as a phone call.

telephony

A phone number per customer

Each of your customers gets their own number and their own assistant, provisioned through the API as part of your onboarding rather than by hand. Inbound, outbound and transfers all behave the same way underneath.

full white-label

A branded portal you resell

Your own domain, logo and payment flow, with your customers managing their own agents in an interface that is yours. You set the price, we invoice you for the platform.

Multi-tenant mechanics

Built for one account per customer, not one account total

The part that breaks home-made integrations is tenancy. It is the part we designed for, because our own partners resell to hundreds of end customers.

An organisation per customer

Data, numbers, assistants and usage are partitioned per organisation, so one customer can never see another's calls.

A parent that covers the children

Your organisation can carry usage for the customer organisations underneath it, with a pricing multiplier per organisation. You choose your own margin.

Configuration decided at call time

Instead of storing a saved assistant per customer, you can answer a webhook the moment a call arrives and return the configuration for that specific caller. Onboarding a customer becomes a row in your database, not a manual setup.

Usage you can bill on

Per-organisation usage is what your own invoicing runs on, whether you charge per minute, per seat, or roll it into a plan.

assistant-request.ts
// your backend answers the assistant-request webhook
// (fires before the call is answered, 5s to respond)
export async function POST(req: Request) {
  const { message } = await req.json()

  // look the tenant up by the number that was dialled
  const tenant = await db.tenant.findByNumber(message.phone_number.number)

  return Response.json({
    assistant: {
      first_message: tenant.greeting,
      llm_config: {
        messages: [{ role: "system", content: tenant.prompt }],
        tools: tenant.tools,
      },
    },
  })
}

The escalation that matters

When your customer calls you, you need someone to call

This is the real reason embedding voice is different from embedding a payment widget. A phone line fails in ways your logs cannot explain, and your customer will ask you, not your vendor. If the layer underneath you is a black box, your answer is a support ticket and a shrug.

You escalate to the people who wrote the orchestration layer, not a queue
Telephony problems get read in the SIP-trunk logs, and you get the explanation
We fail closed with a readable reason, so your own support has something to work with
No lock-in: the stack is transferable and on-premise deployment exists, both under a licence we agree with you

Questions

What software companies ask first

Will my customers know you exist?

Not unless you tell them. In the white-label setup the domain, logo and payment flow are yours, and in an embedded setup the voice simply appears inside your product. What we do ask is that the agent itself discloses it is an AI to the caller, because from 2 August 2026 EU transparency rules require it.

How long does an integration take?

A working in-app voice call is a small piece of work: request a token, open a session. The real timeline is set by your own tool calls, the lookups and actions the agent has to perform in your systems, and that is exactly the part we build alongside your team.

Can I use my own model providers?

Yes. Bring your own keys and your providers bill you directly at your rates, with the orchestration fee unchanged. Or run on our platform keys, where model usage is passed through at cost.

What does it cost me per customer?

Orchestration is € 0,07 per minute, plus model usage and telephony. You set your own price to your customers, so the margin is yours to decide. Most partners charge per minute with a plan on top.

What happens to my customers if we stop working together?

The stack can be handed over, and on-premise deployment is available, both under a licence agreed up front rather than something you flip on later. That exit is deliberate: an embedder cannot responsibly build a product line on infrastructure it has no way out of, and we would rather win the renewal than hold the door shut.

Is this the same as your agency programme?

It overlaps but the emphasis differs. The agency route is about selling voice as a service to end clients. This page is about voice becoming a feature of software you already sell, where the integration lives inside your product and your onboarding.

Tell us where voice belongs in your product

Thirty minutes with the people who own the layer. We map the integration, the tenancy model and the tool calls, and tell you honestly what the first version should be.