Big When Needed.
Off When Not.
Critical workflows don't run at the same size all day. Start the machine each job needs, run your business rules on data in place, and stop it, or let it stop itself when idle. A bigger machine never changes the answer.
rules v12 · one rule setProven in production
Always On.
Or On Demand.
Most rule-heavy workloads are spiky: a quick test in the morning, a heavy production run in the afternoon, nothing overnight. Sizing one cluster for the peak and leaving it on means paying for idle time. Prometheux sizes compute to the job.
Right-Size
Every Job
Choose a machine from the catalog for the work in front of you. Use a small one to build and test rules, switch to a large one for the production run, and turn it off afterwards. Compute is used while a machine runs, not while it sits idle.
Machine Catalog
Pick the size each task needs, from quick queries to heavy, multi-hop production runs.
Start and Stop
Power a machine on when the work starts and off when it ends. Its status is visible as it starts, runs and stops.
Auto-Stop When Idle
Set an idle timeout and the machine stops itself, so a forgotten session doesn't keep running.
Agents Can Manage It
Agents and scripts can list, start and stop machines through MCP and the CLI. Our MCP tools are labelled read-only or destructive so your client can ask first, and the CLI asks before applying changes.
Stop paying for idle clusters. Give heavy jobs the power they need. Keep development cheap and fast.
One Workspace.
Any Machine.
Machines come and go; your workspace doesn't. Every Prometheux-managed machine you start connects to the same rules, files and results, so moving a job from a small machine to a large one is a click, not a migration. Source data stays in your systems.
Same Storage
Small and large Prometheux-managed machines read and write the same workspace, so results from one are there for the next.
Private by Design
Each user gets their own pod, network rules and storage folder, with cloud credentials limited to that folder, even when compute is shared.
Same Rules, Any Size
Business rules don't change when the machine does, so a supply-chain risk check returns the same plants on small or large.
Data Stays in Place
Machines query your warehouses and databases where the data lives. Starting a bigger one never means a new copy.
Scale up without moving a file. Keep every team's data private. Get the same answer at any size.
From Idle to
Running and Back
Pick a Machine
Choose a size from the catalog for the job: small to build and test, large for the production run.
Start It
Power it on from the platform, the CLI or an agent. It connects to your workspace and your data sources.
Run Your Rules
Run concepts and workflows as usual. Every result keeps its lineage back to rules and source data.
Stop It
Turn it off when the work is done, or let the idle timeout do it. Your rules and results stay in the workspace.
Built for Spiky
Workloads
Cheap Dev and Test
Build and test rules on a small machine all day, and only reach for a large one when a full run needs it.
Month-End Runs
Start a large machine for the regulatory or month-end batch, then switch it off until next time.
Disruption Checks
When Kora Steel is disrupted, start a machine, find every plant at risk across Lyon, Turin and Gdańsk, and stop it.
Agent Workloads
Agents start the compute a task needs over MCP and let idle timeouts close it, instead of keeping clusters warm.
Questions,
Answered
What is elastic compute in Prometheux?
You choose a machine from a catalog for each job, start it when the work begins and stop it when it ends, or let an idle timeout stop it for you. Your rules, files and results stay in your workspace between runs.
Do I lose anything when a machine stops?
No. Machines connect to your workspace and to your data where it lives. Stopping one ends the compute session; the rules, files and results remain, and the next machine you start picks them up.
Is my data private if compute is shared?
Yes. Each user gets their own pod, network rules and storage folder, with cloud credentials limited to that folder, and your source data stays in your own systems.
Why not a serverless platform that scales on its own?
Those scale general compute. Prometheux sizes compute for rule workloads, keeps one private workspace across every managed machine, and returns the same result on any size, with lineage back to rules and source data.
Can we use our own Databricks or Kubernetes compute instead?
Yes. Prometheux can run as a native app on Databricks or on your Databricks clusters, and on self-managed Kubernetes clusters, where executor count, cores, memory and dynamic allocation are configurable, with GPUs in GPU-enabled environments. See Portable Cloud for every deployment option.
Related Capabilities
Pay for Work.
Not Idle Time.
Bring a workload that runs hot some days and cold the rest. We'll show you how it runs on right-sized machines.
