Cognitive tooling · Architecture

The architecture that makescognitive tooling work.

A holon is something whole on its own and a part of something bigger. Holonic is the open source architecture built on that idea - the one every Fathym product runs on, and one you can build your own systems with.

Why it compounds

The composition is the cognition.

Holonic is the architecture that makes the Fathym family work as cognitive tooling.

fAI compounds because workbenches are holons. OpenX coordinates because workshops are holons. PowerAI composes because platforms are holons.

Why it matters

One shape, at every scale.

Every entity is a holon

Autonomous - you can reason about it in isolation - and composable into a larger whole via typed inputs and outputs. It's both at once: whole and part.

A bounded, 4-tier holarchy

Workbench → Workshop → Studio → Platform. The same shape at every tier, so the cognitive load of learning a new holon stays small - and the bound is what keeps the whole thing stable.

Composability keeps it legible

You reason in a handful of meaningful chunks, not dozens of moving pieces. Composable holons keep the working-memory load small - and leave more attention for the calls that need human judgment.

Where the AI/human boundary lives

Teams encode their own working patterns as their own holons - versionable, auditable, owned. That is the line between what AI runs and what humans decide.

Open and proven

The architecture you build on.

Holonic isn't a tier or a SaaS offering; it's the architecture every Fathym product crystallizes from. And it's already shipping - the code is public, the architecture is proven.

Open today

@fathym/steward

The core holonic architecture - released and consumable on JSR. Workbench, tool, capability, operation, vault, all the same holon shape.
Open today

@fathym/power-ai

The engine behind fAI - holonic top to bottom, released and consumable.
Open today

@fathym/cli

The command line for the whole stack - holonic, released, consumable.

Coming to holonic.fathym.com: @fathym/holonic - where steward's core consolidates into one open package - plus a training corpus (old-to-new code pairs) and a public MCP server for holonic review and scaffolding.

Architecture

The same shape, at every level.

A holon has two faces - whole and part. A Workshop contains a Workbench but never erases it: a larger whole that keeps the smaller one intact. That's what lets a system grow without its parts losing their meaning.

The 4-tier bound - Workbench, Workshop, Studio, Platform - is what keeps the holarchy stable: deep enough to model any organization, bounded enough to stay legible.

The code is public; what compounds is the architecture. One shape runs from the engine behind fAI to the substrate under OpenIndustrial - so every product, and every new holon, is built the same way as the last.

One open-source architecture under every product - and each new holon makes the next one easier to build.

In plain terms

Every part of the Fathym family is a holon - whole on its own and composable into something larger. That's what makes a workbench compound, a workshop coordinate, and a platform bring it all together without changing what each one is. Humans and AI work with the same composable patterns, which means you can reason about big systems without big context.

Go deeper.

The concepts, patterns, and guides for building on the holonic architecture.