Run AI voice agents on your own infrastructure

An on-premise AI voice agent license is a one-time license to deploy and operate a voice AI platform inside your own server or cloud account, so every call recording, transcript and customer record stays within your infrastructure instead of a vendor's multi-tenant SaaS. In practice it means the voice agent answering your phone runs on infrastructure you control, and no third-party platform ever holds the recording of a customer call.

Yes — on-premise, self-hosted and private-cloud describe the same arrangement here, and buyers use the terms interchangeably. What matters is the tenancy: the platform runs in an account you own, whether that is your own hardware, a dedicated VPC in AWS, Azure or GCP, or a data centre you already operate. We deploy into whichever of those your security team has already approved.

No. Once deployed, the application, database, object storage, recordings, transcripts and user records all live inside your environment. There is nothing sending call data back to us, and we hold no standing production access. Anything you choose to share while raising a support issue is shared deliberately by you, and nothing moves without that.

No. This license covers deploying and operating the platform inside your own infrastructure for your own business; it does not include source code, and it does not include the right to resell, sublicense or host the platform for anyone else. If holding the code and selling the platform on to your own clients is what you are after, that is our white-label license instead.

The on-premise license is for running the platform for your own organisation: your calls, your data, your locations, no resale. The white-label license is for selling it — it includes the full source code and the right to rebrand the platform and operate client accounts under your own name. Same platform, different rights, different buyer.

Yes, and there is no per-seat, per-location or per-tenant charge for doing so. Each site or team gets its own numbers, greeting, business hours, routing rules and agent configuration inside the deployment you already run. The only thing that grows as you add locations is infrastructure and usage, billed to you at provider rates.

Substantially, because it removes the questions those reviews usually fail on. Security assessments rarely stall over encryption — they stall over where data physically lives and who else can reach it. Running the platform inside your own environment puts call recordings and transcripts under the controls, safeguards and agreements you have already had audited, and makes GDPR export and erasure requests queries you run yourself rather than tickets you file with a vendor.

A cloud account or server environment you control, plus provider accounts for telephony and for speech and language models. The exact sizing depends on concurrent call volume, and we work it out with your team during the architecture review before anything is committed, so there are no surprises on your first infrastructure invoice.

Most deployments go live within days of the agreement and infrastructure access being in place. Where it takes longer, the reason is almost always on the approval side — a security review, a network change or a carrier porting window — rather than the deployment itself, and we will flag which of those apply to you on the fit-check call.

It is a one-time license, quoted once against your deployment scope, with no per-minute platform fee and no per-seat or per-location charge. What varies between deployments is the environment, the integration work and how much of the deployment you would like us to drive — so we quote it accurately for your situation rather than publish a figure that would be wrong for almost everyone reading it. Telephony and model usage are billed to you directly by those providers at their rates, with no margin taken by us.

The difference is structural rather than a discount. Per-minute platform pricing rises every time the agent answers another call, so the better the system works the more the platform costs. An on-premise license fixes the platform line permanently and leaves only genuine usage — carrier minutes and model inference — variable. That is why the case is strongest for high-volume operations and weakest for low, occasional volumes, where our hosted plans are the honest recommendation.

Yes, and being inside your own network makes it easier rather than harder. There are documented REST APIs and webhooks, plus HubSpot, Salesforce, Google Calendar, Outlook and Calendly integrations out of the box. Because the platform sits inside your perimeter, it can reach internal databases, practice-management systems and line-of-business tools that you would never expose to an external SaaS vendor.

Updates and engineering support are covered by an ongoing term agreed at deployment, so platform improvements — model, voice and telephony changes included — reach your instance on a schedule you control rather than being pushed at you mid-shift. You choose when to apply them, and if you end the term your instance keeps running exactly as deployed.

Your deployment is unaffected, which is much of the point. The platform runs on your infrastructure under a license already granted, so there is no account for anyone to reprice, throttle or switch off. Your only exposure to us is the optional support and update term, and ending it leaves the platform running.

We cover your call volumes, the environment you want to deploy into, your compliance obligations and which integrations matter, then walk you through the platform and the architecture so your technical people can interrogate it properly. If on-premise is the wrong shape for you, we will tell you on that call. There is no obligation at any stage.

It follows from one question: who is allowed to hold the call recording. If a shared multi-tenant platform holding customer audio is acceptable to your reviewers, SaaS is simplest. If it is not, a private tenant narrows the blast radius but the vendor still holds the data. Self-hosted is the only model where recordings, transcripts and caller PII never leave infrastructure you control, which is why regulated buyers usually end up there regardless of where they started.

Role-based access control, single sign-on, audit logging of who accessed which recording, configurable retention and deletion, data residency you can evidence, and an integration model that does not require broad standing credentials into your core systems. Ask how each is evidenced to an auditor rather than whether it exists.

Because you choose the cloud account, region and data centre, residency stops being a vendor promise and becomes a line in your own infrastructure inventory. For organisations operating across regions, separate deployments can be run per region so data never crosses a boundary it should not.

We do not publish compliance certifications as marketing claims, and we would encourage the same scepticism toward any vendor that does without showing you the report and its scope. What an on-premise deployment does is remove us from the data path entirely: the controls, the residency and the retention become yours, inside an environment your auditors already assess.

Ask where audio is processed and stored, which sub-processors touch it, how long it is retained by default, who inside the vendor can access a recording, how access is logged, what happens to your data if you leave, and whether any model provider retains prompts or audio for training. The answers to those seven questions separate platforms far more reliably than a certification badge.

Concurrency is provisioned against your measured peak rather than an average, and because the deployment runs on your infrastructure you control the headroom rather than sharing a pool with other tenants. Size it from your busiest hour, not your monthly total, because the busiest hour is what fails.

Yes, and that is precisely what this license is. The platform is deployed inside your own cloud account, VPC or data centre, answering your phone lines, with no per-minute platform fee and no call data leaving your environment.