On-premise deployment — installed on-site, remotely

What you need on-site

MPX ENGINE runs on a GPU-equipped server, installed on your premises and connected to your local network. Two options:

• You already have a GPU server: we validate its configuration together before installation.
• You don't have one: MPX supplies a turnkey server, sized for your workload, quoted on request.

On the network side, nothing exotic: your existing LAN, plus temporary remote access (SSH) opened for the installation, which your IT team controls and closes afterwards.

A 100% remote installation

No IT project, no consultants on the shop floor. Deployment is done entirely remotely: controlled SSH access for the technical install, video call for configuration with your team.

The whole platform is containerized (Docker): database, extraction engine, prediction engine, interface — one coherent set, installed as a block, manageable by your IT team with their usual tools. Configuration covers your users, their roles and your sites.

Your work never goes through the internet

None of your work goes through the internet. The AI models — BOM extraction, machining-time prediction — execute locally, on your server. No call to a cloud API, ever.

One exception, and we state it plainly: the license heartbeat. Periodically, the server sends a single outbound request to MPX to validate the license and the list of active modules. That is all it carries: no drawing, no BOM, no price, no technical data. Your IT team can inspect it, log it in their firewalls, and confirm it is the machine's only permanent outbound connection.

Your data never leaves

Your customers' drawings, BOMs, 3D files, quoting history, margins: everything lives on your server, within your walls. No cloud copy, no third-party storage, no training data going back to the vendor.

It is compatible with your customers' confidentiality requirements — including on the most sensitive programs.

Updates

Updates are prepared by MPX and applied remotely, on a slot agreed with you — never behind your back. The containerized architecture makes it possible to roll back to the previous version if needed. No surprise downtime: you know what changes, and when.

Who does what

The split is clear: MPX is responsible for the application — how it runs, its updates, its support. You stay in charge of your infrastructure: the server if it is yours, the network, the access rights, and how the platform fits your IT policy.

Your IT team keeps control of what matters to them: where the machine runs, who can reach it, what leaves the network. Everything we install is documented — architecture, containers, configuration — and handed over to your team. Nothing is opaque.

Your team trained on your own data

Installation wraps up with a live video training session with your users — estimators, methods engineers — on their own parts and their own RFQ packages. We take the time it takes: the session ends when your team can run with it, not when the clock runs out. Then enhanced support through the first 30 days after go-live, so the platform takes root in your daily work, not in a drawer.