konnyaku.
THE UNDERSTANDING LAYERHuman first. By design.

You shouldn’t have to
speak like a machine.

Konnyaku helps machines
understand what you mean.

An understanding layer between humans and AI agents.
Connecting human intent, business context, and the language of systems.

See how it works
Less machine-speak. More meaning.
01 /A FAMILIAR PROBLEM

We have always translated
ourselves for computers.

CommandsFormsMenusKeywordsSchemasPrompts

Different interfaces. The same compromise.
We learn the system’s language before it can understand ours.

AI is changing that. But as agents move from answering questions to taking action, words alone aren’t enough.

Understanding words≠understanding intent.
02 /FROM WORDS TO MEANING

One sentence.
More than words.

A natural request carries a lot of unstated meaning.
Understanding starts by making it explicit.

A HUMAN SAYSIllustrative example
“Can you take care of this customer?
They seem a little unhappy.”
WHAT WE KNOWNo customer record or conversation has been provided yet.
Konnyaku · making the meaning explicit
01 / INTENT

What are you really asking for?

The likely goal is to address a customer’s concern. The request doesn’t yet tell us which customer, what happened, or what action is wanted.

WORKING INTERPRETATION

Understand the concern and prepare an appropriate follow-up.

An interpretation to confirm — not a fact to assume.

A possible structured intent

A conceptual handoff for an agent. Missing information and approval requirements stay visible.

{
  "goal": "address_customer_concern",
  "customer_id": null,
  "evidence": [],
  "needs_clarification": ["customer", "issue", "desired_action"],
  "proposed_next_step": "ask_for_customer_and_context",
  "external_actions": "require_human_approval"
}

A guided concept demonstration, not a live AI service. No accounts are connected and no actions are executed.

03 /THE LAYER IN BETWEEN

Human on one side.
Understood on the other.

A place for intent, context, and boundaries to become clear before an agent acts.

THE HUMAN SIDE

Your words. Your world.

Natural requests · Business context · Unspoken expectations

konnyaku.
The Understanding LayerIntent · Context · Meaning · Boundaries
THE SYSTEM SIDE

Something machines can work with.

Agents · APIs · Enterprise systems

An architectural direction, not a list of shipped integrations.
We’re exploring how this layer should work in real enterprise workflows.

THE BELIEF BEHIND THE LAYER

Machines should learn
to understand us.

Not the other way around.

04 /A SMALL NAME. A BIG IDEA.

Why Konnyaku?

A childhood imagination.
A very human ambition.

The name draws on the familiar idea of “translation konnyaku”: a small piece of food that makes understanding one another effortless.

That simple imagination stayed with us. What if technology could understand us without asking us to become technologists first?

Konnyaku brings that question to the space between human intent and machine execution.

BORN FROM A SIMPLE QUESTION

What if being understood
came naturally?

KONNYAKU

Translate meaning,
not words.

The understanding layer
between humans and AI agents.

Explore the layer