There are LLMs.
There are World Models.
There is no Operational Model.
Between your data existing and something actually happening, there is a layer nobody built. Declare your operating model and the policies that govern it once — Inferal runs them continuously and activates agents the moment conditions match.
The gap
The stack between "data exists" and "something useful happens" is improvised in every company, by hand, over and over:
Every company builds this. Nobody calls it anything, because the category doesn't exist yet. That's the layer we're building.
How it works
Not a workflow you draw. Not code paths you maintain. A description of your business and the rules that govern it, running continuously against live data.
The entities and relationships that describe your domain. You don't write it from scratch — pre-mined models ship by industry, and you customise from there.
accounts · transactions · devices · sessions
Standing orders expressed in business terms, not code paths. They run continuously and fire the moment conditions match. Inferal can suggest them from your own data.
when X holds across Y for Z → freeze, review
Connects to your stack, keeps the model in sync with reality, matches policies against it, activates agents with the context they need, and traces every decision.
your cloud · on-prem · Kubernetes
Getting started
The hard part of a system like this has always been the first day: describing your domain and writing rules before anything works. So Inferal does that part first. Connect a source and it reads what's there, tells you what it found, and proposes policies calibrated to your numbers — thresholds derived from your margins and volumes, not defaults. You approve the ones you want.
Boundaries
The category is new, so the fastest way to explain it is by elimination.
| Not a database | Databases hold everything and do nothing with it. |
| Not a workflow engine | Workflows follow fixed tracks. Policies respond to real conditions. |
| Not an ETL pipeline | ETL moves bytes. Inferal activates behaviour. |
| Not monitoring | Monitoring wakes a human at 3am. Inferal activates an agent. |
| Not an AI wrapper | This is infrastructure, not a model with a prompt in front of it. |
Products
Where operating models and their policies live and run. Matches continuously, activates agents, traces every decision back to the policy and the data that triggered it.
Turns unreliable external APIs — Meta, Google, Shopify, WooCommerce — into continuously fresh streams. Policies can only fire on data the system can actually see.
Living understanding of complex codebases: maps, terminology, reading paths, drift detection. Context for anything that has to reason about code.
Design partner program
If your business logic is spread across cron jobs, queues and middleware you'd rather not own, we'd like to talk. Design partners help shape the operating model libraries for their domain.