Amorfs gives AI agents knowledge they can interrogate directly.
Today’s AI is exceptionally good at finding and generating information.
But when the answer depends on having the exact entity,
relationship or timeframe and where it came from, then things becomes much harder.
Amorfs is a knowledge format and query engine designed to preserve that context,
with identity, relationships, time and source baked in.
It turns complex datasets into structured, connected facts that
people and AI can understand and use.
An AI agent can search millions of documents in seconds.
But some questions aren’t so easy to answer:
These aren’t simply search questions.
They are questions about identity, relationships, time and provenance.
Traditional databases can answer them, but agents often have to
guess what fields mean, invent the joins, and pick the right query.
Even when a search returns relevant information,
that is not the same as having the exact record.
Search comes with no guarantee of completeness either.
Amorfs puts meaning and context
alongside the facts themselves.
The idea
Give AI the facts, not just the text.
Amorfs represents knowledge as connected facts rather than isolated
pieces of information.
The identity of something, its relationships to other things, when those
relationships were true, and the sources supporting them all travel with
the knowledge.
That means an agent doesn’t have to reconstruct the whole story from
fragments every time it needs an answer.
Identity
Know precisely what it is.
Relationships
Know how it is connected.
Time
Know when it was true.
Source
Know where it came from.
The context is part of the knowledge.
Three layers
Amorfs
The knowledge representation and query model.
Knowledge Packs
Bounded, domain-specific knowledge built using Amorfs.
First Cognition
The company that creates, maintains and deploys Knowledge Packs for AI agents.
The difference
Search finds text that looks relevant. Amorfs links facts and what they mean.
AI shouldn’t be piecing together disconnected fragments of text to figure out
correct identities, connections, and historical accuracy. Specific context
should travel with the knowledge itself.
That’s the difference between retrieving information and working with
knowledge an agent can interrogate directly.
The format
Each fact keeps its meaning.
Knowledge is written down in a text format that is intuitive and human readable
so it is easy to understand and validate.
A person can read that. An agent can use it. Neither has to reconstruct
the story from scraps of text.
What that changes
The exact record, not a near match.
Ask for a specific asset, company or legal matter, and that is
what comes back. Not something that looks similar.
The date is part of the answer.
“Installed here between these dates” is built into
the fact. Ask the same question as of a different date, and the
answer changes with it.
The source sits with the fact.
Which notice authorised the work, which change number approved
the revision — that evidence travels with the answer.
The agent does not have to assemble it afterwards.
Only the knowledge that is needed.
The agent is given the facts for that task, not access to the
whole system. Answers stay small enough to use.
If the answer is incomplete, it says so.
If 6 of 29 records come back, that is stated. Search returns
relevant passages and does not tell you what it missed.
The context travels with the knowledge.
Try it
Two ways in.
Play with the format
Type or paste Amorfs and see the same knowledge as a card, table, or graph. No installation. No account.
Imagine an industrial plant represented as knowledge an agent can
interrogate directly.
Not just a list of equipment. But the equipment, its location, the
materials installed, the relationships between components, the relevant
dates and the history surrounding them.
Our synthetic refinery-unit demonstration contains approximately
258,000 facts.
An agent can interrogate that knowledge without having to reconstruct
the entire picture from disconnected documents or complex API calls
and queries.
One plant. Hundreds of thousands of facts. Connected knowledge.
The next generation of AI agents will do more than generate text.
They will investigate, reason, make decisions and take actions.
For that to work reliably, an agent needs more than a large collection
of documents. It needs to understand:
What is this?
What is it connected to?
When was that true?
What supports the claim?
Amorfs is knowledge infrastructure for agents: a structured way to
represent that knowledge so they can interrogate it without having to
reconstruct the context from scratch.
Knowledge Packs
From knowledge format to working AI infrastructure.
Amorfs is the knowledge representation. What you buy is a Knowledge
Pack — a bounded set of records an agent can hold locally —
and a four-week evaluation on one domain.
A pack can hold the knowledge for a particular domain without giving
an agent unrestricted access to the production system.
We’re looking for organisations with complex, connected data where an
AI agent needs to get the facts exactly right.
Examples include:
Asset and maintenance data
Master data
Ownership and KYC
Legal records
Other connected enterprise records
The first step in becoming a design partner is requesting a demonstration.
The evaluation is paid and runs four weeks: one bounded domain, one
controlled dataset, twenty real questions with known answers, and a
written scorecard against the current approach.
One domain. One controlled dataset. Real questions with known answers.
