The open suite for process plants.

Historian, analytics, alarms, dashboards, process models and plant documents in one place, with an AI assistant that checks the results with you.

Try the live demo Install free Source on GitHub

01 · Open

All of it open, and yours to extend.

Every line of code and every method is open. Nothing is a black box: each result says how it was reached, in blocks and Python you can read, change and share.

  • Apache-2.0Use it, change it, ship it inside your own products. No paid edition, no locked feature.
  • Open methodsThe performance and data quality methods are open and written in Python, from public literature, so you can check, cite and improve them.
  • A complete suiteApps for data, analytics, alarms, dashboards, models and documents, ready to install and configure, sharing the same people, assets, signals and look.
  • ExpandableNew apps for your process, your research or your client are created from a template and slot in beside the official ones.

02 · Secure

Built for the plant's IT and OT review.

Plants install nothing their security teams have not approved. PlantCraft is built for that review from the first line, and a two-page security guide maps it to the Purdue model and IEC 62443.

  • On your own infrastructureInstalls on your servers and inside your network, from one signed bundle, even with no internet. No telemetry, licence call or outside service.
  • Read-only to the plantReads OPC UA, PI Web API, MQTT, Modbus TCP, SQL and CSV. It never writes to a DCS, PLC or historian.
  • Protected by defaultOne HTTPS port with sign-in and an audit log, calculations isolated with no network or secrets, and a backup before every update.
  • AI you configureOff until an administrator turns it on. Connect the provider you choose, or a model your company already runs. The assistant reads only what the person asking may read.

03 · LLM-first

An assistant that finds the problems with you.

Instead of searching through results and charts, ask. The assistant checks the calculations and the results, points to what needs attention and suggests what to do, so your time goes to understanding the problem and planning the fix with your team.

  • Checks the resultsIt reads the plant's data, the results of every app, its documents and history, and answers with charts and links to its sources.
  • Suggests actionsIt says what is wrong, why, and what to do first. A person always confirms.
  • Helps you build, tooIt proposes new blocks, algorithms, alarm rules and models with their fields filled in, for you to review.
  • Your own AI, tooCopy any block, model or the plant map as a prompt, or let your assistant read the plant through a read-only MCP server.

Why did this alarm fire?

What should we fix first this week?

What changed in the furnace since Monday?

Help me create a block that detects a sticky valve.

Help me design a model of this heat exchanger.

04 · Fast

Value in days, not months.

No project before the first result. Install it, point it at the plant's data, and the first findings arrive the same day. Measured in our tests on a 4 vCPU, 16 GB server with 8,000 signals; a larger server does more.

under 1 hour
to install from the offline bundle
0.2 s
to run one algorithm on one asset
21 ms
for a 24-hour trend of one signal (p95)
29,000+
values a second written by the Historian

Sources: Install PlantCraft and Sizing the server. Run the same bench on your own machine.

The suite

One platform. Every plant tool.

Each app runs in its own container and shares the same people, plant assets and look. Install the ones you need.

Desk

Your day

What needs you, actions, calendar, to-do, notes and the team on shift.

Historian

Data

Collects and archives the plant's signals from OPC UA, PI Web API, MQTT, Modbus TCP, SQL and CSV, read-only to the plant.

Analytics

Signals and assets

Algorithms of blocks that watch every signal and asset and say what is wrong and why.

Alarms

Operations

Alarms and alerts, shelving, history and alarm performance.

View

Dashboards and diagrams

Dashboards, process diagrams, trends and scheduled reports, shared with the shift.

Models

Modelling

Block-diagram models with Modelica underneath: simulations, studies, optimization and FMUs.

Documents

Plant knowledge

Manuals, control narratives and datasheets, searched offline and cited by the assistant.

Also inside: Jupyter notebooks with the plant's data, in the browser or on your own computer, and a simulated plant to try everything safely before connecting a real one.

Install

One bundle, one command.

Made for networks without internet: copy the bundle to the server, check its signature, run the installer. It sets up the platform, the apps, the demo plant if you want it, and an administrator account.

Linux for plant servers; macOS and Windows with Docker Desktop for trials. The install guide · Requirements

# Linux with Docker; the bundle from the releases page
tar -xzf processcraft-0.5.0-linux-amd64-bundle.tar.gz
cd processcraft-0.5.0-linux-amd64-bundle
./install.sh --all-apps --demo
# open https://<server>/ and sign in as admin

Talk to us

Tell us about your plant.

A pilot on your unit, help to install it, training for your team, an app for your process, or a question the documentation does not answer: write to us and we answer personally.

Try the live demo
Live demo