About

Where compute becomes action.

New models arrive faster than any team can absorb them. ATP Token puts them all behind one governed key: who used which model, how many credits it took, and who changed which setting — all on record.

Ecosystem

Token Factory: from data center to every request

KONST treats compute as production equipment and tokens as the output. The group runs three layers: building and operating GPU capacity, slicing it into usage-based units, and governing how enterprises spend it. ATP Token is the top layer.

  1. KONST
    01

    KONST group · compute supply

    Konstra AI

    Data center siting, power and cooling, GPU cluster tuning, acceptance and ongoing operations across Asia.

    Compute

  2. Glows.ai
    02

    Sister company · conversion layer

    Glows.ai

    Splits cluster compute into small units so teams pay for what they use, billed by the second.

    Usage units

  3. Horizon AI
    03

    Our company · governance and delivery

    Horizon AI

    Builds ATP Token: permissions, usage and cost management for every request, from the first proof of concept onward.

    Governed tokens

Action · Your teams

Company

Built by Horizon AI

ATP Token is built by Horizon AI, an enterprise AI adoption and governance company and a member of the KONST group. Horizon AI works with enterprise groups and large organizations on AI rollouts that span departments, existing systems and long-term governance. ATP Token is the infrastructure layer of that practice: model access, usage governance, and billing consolidation all work at this layer.

Meet Horizon AI
Horizon AICompany
Horizon AI
  • Legal nameHorizon AI Co., Ltd.
  • GroupKONST
  • RoleAdoption & governance
  • ServesEnterprise groups and large organizations
Builds ATP Tokenhorizon-ai.ai
What We Build

What we are building

  1. 01

    Governance layer

    Organization, workspace, project, key: four layers decide credit allocation, model access, budgets and request logs. Access follows structure, not individuals.

  2. 02

    Billing layer

    Usage across providers is logged and attributed per request. At scale, many models can sit under one contract and one bill, planned around actual usage.

  3. Roadmap

    Enterprise AI infrastructure

    Governance goes beyond models. The platform extends toward agent orchestration and a Tool Registry, so tools and workflows inherit the same permissions and billing.

Name

What's in the name

The full name is Agent Token Petrichor. Two of the three words describe how the platform works — who is spending, and what it is denominated in. The third is simply where the name came from.

  1. AAgent

    Who spends

    Requests no longer come from a person at a keyboard. They come from agents that loop, retry, and call tools on their own — which is exactly why spending needs a structure that does not depend on someone remembering to check.

  2. TToken

    What it is denominated in

    Text models are metered in tokens; other models use their published image, second or character rates. ATP records each request under its project and settles usage in credits.

  3. PPetrichor

    Where the name came from

    Petrichor is a rare English word, coined from Greek roots (petra + ichor). Here it is only etymology — where the letters came from — not a product metaphor and not part of the brand promise. What the product does is plain: compute becomes action that is attributed and accounted for. The full name lives on in our legal entity and in older material; the product is ATP Token.

The initials are not a coincidence. In biology, ATP is the cell's energy currency: whatever the source, energy has to be converted into ATP before the body can spend it. Compute works the same way — it has to become tokens before an enterprise can actually use it. That conversion, metered and accounted for, is the layer we work at.

Get started

Bring AI governance and billing under one roof.

Start with one key, or talk to us about your rollout.