KINODEGet in touch

Platform vision

One API. A Fabric of Compute.

Our vision is to make AI workloads accessible through a trusted inference platform powered by heterogeneous, owner-authorized computing resources.

Long-term direction. Not a live service.

  1. Developer Platform

    Future model APIs, inference jobs, usage metering, billing and deployment experience.

    Future vision

  2. Intelligent Execution Fabric

    Eligibility, admission, placement, execution strategy selection and resource protection.

    In development

  3. Compute Supply

    Trusted community devices, enterprise-owned resources and future partner-contributed capacity.

    Future vision

The long-term analogy is distributed compute supply similar to Salad combined with a developer inference experience similar to fal.ai. These are analogies for a business direction. Neither company is a partner or integration, and no feature parity is claimed.

A phased progression

From your devices to a programmable AI compute network.

Conceptual diagram of compute supply widening in four stages around the Kinode fabric. Only the first stage is current work.1234
Conceptual · no external network is operating today
  1. 1
    In progress

    Private Compute Fabric

    Trusted devices, heterogeneous execution, resource governance and physical capacity validation.

  2. 2
    Strategic stage

    Managed Private Inference

    Controlled model endpoints, team or enterprise deployments, reliability and repeatable workloads.

  3. 3
    Strategic stage

    Partner Compute Pools

    Institutional and corporate resource contributions, including sponsored scientific computing.

  4. 4
    Strategic stage

    Kinode Inference Platform

    Permissioned external compute supply, developer-facing model APIs, metering and usage-based business models.

These are strategic stages, not approved delivery commitments. No public Kinode marketplace, pay-per-token service or production model-serving network operates today. Offline, research, sponsored compute and enterprise use cases remain part of the strategy.

Compute resource scenarios

From Idle Hardware to Useful AI Capacity.

  1. Idle
  2. Eligible
  3. Owner Approved
  4. Allocated
  5. Released

Illustrative lifecycle, not measured live states.

Potential resource owners, resources, workloads and governance
Resource OwnerAvailable Computing ResourcesPotential AI WorkloadGovernance Requirement
Independent DeveloperAvailable Computing ResourcesPersonal Mac and gaming PC.Potential AI WorkloadPrivate open-weight model experiments.Governance RequirementOwner sets which devices and how much capacity participate.
Small CompanyAvailable Computing ResourcesApproved off-hours workstations.Potential AI WorkloadInternal batch inference.Governance RequirementCompany approval, contribution windows and internal data rules.
University LaboratoryAvailable Computing ResourcesMixed-generation research computers.Potential AI WorkloadShared research workloads.Governance RequirementLab-level allocation between research groups.
Corporate SponsorAvailable Computing ResourcesAuthorized nighttime compute capacity.Potential AI WorkloadApproved non-sensitive scientific research workloads.Governance RequirementSigned commitments, quotas and release before working hours.

Potential use cases only, not existing customers or committed hardware supply.

Early target users

For the people
building what’s next.

Starting with small, trusted clusters and practical open-weight inference experiments.

01

AI Engineers

Experiment with open-weight models using compatible existing hardware.

02

Early-Stage Startups

Explore more affordable, controlled AI inference infrastructure.

03

Research Teams

Coordinate eligible computing resources across trusted devices and small laboratories.

Trust by design

Your hardware.
Your control.

Device owners authorize capacity and set limits. Admission, leases and release govern how participating resources are used.

AuthorizeAdmitLeaseRelease
Start hereHow the fabric works today Talk to the Founder