Knowledge core

Choose what to trust.
Put it to work.

Your knowledge. Your control.

A neurosymbolic engine.

Context assembled for the task.

The right context.
A useful check.

Connect the source and version.

Apply the relevant conditions.

Show why the work was flagged.

inside knowledge core

Represent meaning.
Assemble for the task.

Knowledge Core is Primal's neurosymbolic knowledge engine.

It represents concepts, context and rules, the assembles knowledge structures applications can use.

Stored in knowledge core

Concepts and rules

Meaning, relationships and source references.

Context for this task

Purpose and scope

The subject, selected sources and purpose of the task.

Knowledge processing

Analyse. Apply rules. Synthesise.

Assemble a knowledge structure for the question or workflow.

used by your app or agent

Capabilities where work happens

Your app or agent uses Knowledge Core as it works.
Trust Core skills guide its development.

Available for inspection

The basis of the result

Sources and recorded processing steps, where captured.

Architecture illustration. A record of the processing steps shows what the knowledge engine did. It does not reveal an LLM's internal reasoning.

This is the symbolic foundation of Primal's neurosymbolic approach.

Neural models help work with language; explicit concepts and rules provide structures applications can use and inspect.

travel and expense policy ・ illustrative workflow

Can this employee book business class?

Several passages mention business class. Only some apply to this employee and this trip.

Application inputs and scope

One employee. One trip.

Employee role

Staff member

Jurisdiction

Canada

Trip type

International business travel

Flight duration

Nine hours

Selected policy

Canadian employee policy, version 3

relevant knowledge core

Which policy applies?

Applicable: Canadian employee travel policy, version 3.

Inapplicable: US employee policy and superseded version 2.

Condition: Business class requires a flight over eight hours and prior manager approval.

finding and explanation

Eligible if prior approval is confirmed.

The flight meets the duration condition.

Manager approval has not been provided.

Inspect the source and condition

Fictional company policies illustrate the workflow. This is not a live integration or a booking authorization. Supported capabilities and integration are confirmed during scoping.

scope and change

One transaction.
A changing set of documents.

Choose what applies.

A due diligence team receives a document set.

The team chooses which documents to rely on.

Put it to work.

Represent its meaning in Knowledge Core.

Prepare answers and check the summary against it.

Update what changes.

An amended agreement arrives.

Decide what it changes, then review affected work.

These documents are definitive for this transaction at this stage. The team decides whether they can be used for other work.

Illustrative workflow. The application must support source selection, document changes and human review.

A manufacturing example: instructions for a shipment

Institutionally sovereign

Your knowledge.
Under your control.

Knowledge Core is designed to run within your infrastructure.

Your organization owns the knowledge it creates.

You define what is accepted for each use case and when that changes.

Manage change.

Keep concepts and rules outside the model.

Update them as sources and meanings change.

Preserve continuity.

Keep your sources and rules independent of the model.

Connect it to new applications.

Inspect the basis.

Inspect sources and recorded processing.

Identify checks made after generation.

Models can interpret language and reason. Explicit knowledge makes selected inputs and processing manageable independently. Operationsl permissions still required verified identity and enforcement in the application.

Where primal fits

Build apps and agents with Knowledge Core.

Knowledge Core provides representation and synthesis for apps and agents.

Trust Core skills guide coding agents in building with it.

Reuse the capabilities across your workflows, with updates and technical support from Primal.

How is this different from RAG or GraphRAG?
Why not use a graph or vector database?
What comes from Primal's research?
Does this guarantee compliance or correct answers?

Expertise, applied.
Across your work.

© 2026 Primal Fusion Inc. · 210–137 Glasgow Street, Office 163, Kitchener, Ontario N2G 4X8, Canada