Your communication.
Your infrastructure.
Your control.

For organizations where language has consequences. Run VerbaPulse in your private cloud, in your own data center, or in a completely isolated network.

Request access
Illustration · fictional example

Hi Sarah, happy to confirm that the fund will deliver at least 8% this year, so you can plan on that.

CriticalFinancial

Unqualified investment return guarantee enabling reliance by recipient.

More careful wording

is targeting an 8% return this year, but results may vary

The writer decides. The record keeps the risk type and the action, and never the text.

Check a draft →

One engine,
four deployment models

Sovereign

Your private cloud, your control

VerbaPulse runs in a private cloud your organization operates, or in your own cloud tenancy. Draft analysis stays on your servers.

Explore Sovereign

Enclave

Intelligence stays inside

The same software in your own data center, on your own GPU hardware. The analysis servers have no outbound route.

Explore Enclave

Black

Nothing goes out

For isolated networks. Signed install media, offline licensing and updates, with no outbound connection from the server.

Explore Black

VerbaPulse Cloud

Start without installing

A cloud edition hosted in the EU, for teams that prefer not to run their own infrastructure.

Explore Cloud

Which one fits?

Your infrastructure decides

If you use Microsoft Entra and allow one outbound Microsoft address, Sovereign or Enclave. If you want no outbound connection from the server, Black.

Compare

Draft text stays on your analysis servers

In all three private editions the model and the analysis run on servers under your control. The writer always decides.

Model inside your organization

A US-origin open-weight model under Apache 2.0. No inference is sent to third parties.

Your own database

Event records stay with you. Draft text is not stored by default.

Signed license

A signed license file verified offline. No internet connection is needed to validate it.

Measured capacity

Measured on a single NVIDIA H100 PCIe 80 GB card with the full stack running: about 400 users per card at a 12-second p95, assuming 6 checks per user in the peak hour.

Size your deployment
VerbaPulse EnterpriseCapacity measurementH100 PCIe 80 GB, one card
≈400users / GPU12 sp95 response

We plan the deployment with your team

We start with a technical discussion about your network, identity provider and user count. Your IT team installs with Docker Compose and keeps operations in house.

Plan a conversation

Six topics your security team will ask about

Covered in detail on each edition's page.

Deployment

Docker Compose, deployable by your IT team.

Hardware

About 400 users per GPU card. p95 response: 12 seconds.

Model

Western open-weight model. No inference sent to third parties.

Identity

Microsoft Entra on Sovereign and Enclave; activation tokens and your internal identity provider on Black.

Audit

Time-stamped event and administrator records, stored within your organization.

License

Signed offline license, no internet connection required.