The Space Hardware Cloud

Real spacecraft hardware. Accessible like cloud infrastructure.

CubeSTEM is building the Space Hardware Cloud. TwinLink gives engineering teams controlled access to spacecraft hardware owned by laboratories and manufacturers — through one interface, while owners retain access and safety control.

TwinLink — One interface. Many spacecraft systems.

3 min · interactive · hardware execution simulated

Why access breaks

Software moves quickly. Physical hardware still arrives late.

Spacecraft software teams often wait for purchased hardware, custom interfaces, and a complete local bench before meaningful physical validation can begin.

  1. 01

    Interfaces are fragmented

    Every device, vendor, and laboratory introduces another integration path before engineering work can start.

  2. 02

    Access comes too late

    Candidate hardware is difficult to evaluate while architecture and software decisions are still flexible.

  3. 03

    Owners need a safe way to participate

    Laboratories and manufacturers need selective exposure, scoped access, and local authority — not an unrestricted internet connection.

The platform

TwinLink is the control path — not the hardware.

TwinLink sits between engineering applications and heterogeneous spacecraft resources. It standardizes access while keeping integration and physical authority on the owner's side.

One interface
Applications work with TwinLink's common resource and capability model instead of raw device protocols.
Owner-side boundary
The Bridge and Connector stay beside the resource; the browser never becomes a direct hardware channel.
Local authority
Cloud policy can be stricter, but it cannot weaken the owner's final local safety limit.

TwinLink is not the Digital Twin, an Edge box, or a remote desktop. A Digital Twin can be one application-side participant in a future engineering workflow.

Controlled command pathCOMMAND ↓
  1. CONSUME · what a mission team builds on

    01ApplicationMission software or digital engineering tool
  2. 02Client SDKWhat your team writes against — TypeScript or Python
  3. 03TwinLinkAPI, device model, orchestration, access, and evidence
  4. INTEGRATE · what an owner or vendor maps in

    04BridgeOwner-side software — dials out, no inbound port
  5. 05Connector SDKMap your existing protocol — CAN, serial, or a vendor API
  6. 06ConnectorYour mapped interface
  7. 07Local safetyFinal physical authority remains local
  8. 08HardwareOBC · EPS · ADCS · payload · test equipment

Nothing to install and nothing to open.

The Bridge is software you run beside your hardware, on an ordinary Windows or Linux lab PC. It dials out — no inbound port, no appliance to ship, no equipment leaves your site. A CubeSTEM Edge Node exists as an optional appliance and is not on the required path.

↑ TELEMETRY + EVIDENCEReturns through the same controlled path
Illustrative architecture. Physical TwinLink resources are not connected in this website experience.

Composed missions

Bring the hardware you have. Reach the rest through the cloud.

A mission rarely needs an entire bench from one place. You may already own the on-board computer and need someone else's power supply — or a specific attitude-control table, for two weeks in March. A composed mission runs across both at once: your hardware and theirs, through one interface.

YOUR APPLICATION

One Client SDK. One device model. One evidence trail.

TWINLINK

Discovery, leases, authorization, command routing, telemetry, audit

YOUR SITE

Hardware you already own

For example: your OBC

  • Your Bridge — software you run, dialling out
  • Your Connector — your protocol, mapped once
  • Your local safety — yours, and final

You contribute it. You keep it. You can withdraw it at any time.

VENDOR OR LABORATORY SITE

Hardware someone else owns

For example: their EPS

  • Their Bridge — software they run, dialling out
  • Their Connector — their protocol, mapped once
  • Their local safety — theirs, and final

They expose a capability subset to you under an explicit grant: scoped, time-limited, and revocable at any moment.

01Two leases, not one
A session lease is scoped to a single resource. A composed mission holds one lease per resource, each with its own command allowlist and its own expiry. Nothing about one lease is implied by the other.
02Two safety boundaries, both final
Local safety is per site and independent. Cloud policy can be stricter than an owner's limit; it can never be weaker. Neither owner's hard limits can be relaxed by the platform, by the other owner, or by you.
03Two owners, two independent revokes
Either side can withdraw unilaterally, without negotiation. The mission fails closed on that resource — and only on that resource. The rest of the composed bench keeps running.
04One interface
The same SDK, the same device model, the same command paths, the same evidence — regardless of who owns which resource or which country it sits in. Your application does not need to know the difference.

OBC and EPS name the real-world shape of this use case, not shipped device classes — the TwinLink device model covers the resource classes it covers today, and this website is pinned to contracts 0.3.0. The cross-organization grant you can exercise in the product experience is an experience-layer capability, not a TwinLink 0.3.0 API.

How a composed mission works

Choose your path

One platform. Different reasons to enter.

Mission teams use resources. Laboratories and manufacturers own them. Strategic partners inspect how the system can grow.

Trust and safety apply across every path; they are not a separate participant in the network.

01

Mission and software teams

I need hardware access

Discover candidate hardware, begin with simulation, and move toward controlled physical validation through the same engineering interface.

  • Earlier candidate-hardware evaluation
  • Software work before every subsystem is on your bench
  • Measured-versus-model comparison as real resources come online
Explore the mission workflow
02

Universities and laboratories

I operate a laboratory

Start privately, define capabilities and safety limits, then grant selected collaborators access only when you choose.

  • Start private — public exposure is never required
  • Choose what each user group may access
  • Keep safety-critical authority on the laboratory side
Explore the lab-owner workflow
03

Space hardware manufacturers

I build spacecraft hardware

Map an approved product surface into TwinLink without replacing the device firmware, board, or commercial relationship.

  • Existing interfaces mapped through a TwinLink Connector
  • Owner-defined capabilities rather than raw hardware exposure
  • A shorter path from technical interest to technical confidence
Explore the manufacturer workflow
04

Investors, agencies, and strategic partners

I want to build the network

See how controlled integrations can compound into shared engineering infrastructure — without confusing simulated evidence with traction.

  • Interoperability infrastructure rather than a single hardware product
  • Ground engineering first; broader network effects over time
  • Measured proof points added as real hardware is connected
Explore the strategic view

Interactive product experience

Follow one resource from owner setup to mission evidence.

The guided journey runs the real product workflow on the server. Hardware execution is simulated today — and the product is built so the same contract can later be implemented against physical hardware without the workflow you see changing.

INTERACTIVE PRODUCT EXPERIENCE · SIMULATED
Physical TwinLink resources are not connected in this website experience.

01

Owner grants

Scoped telemetry and commands

02

Mission runs

Valid command at 2500 RPM

03

Safety proves

4000 RPM blocked by local max

04

Evidence remains

Attributable result and denial

05

Owner revokes

Future discovery and access close

Server-executed workflowHardware execution: simulated

Trust model

Hardware owners stay in control.

TwinLink is not an open pipe onto laboratory equipment. Owners define what is remotely exposed. Remote access is designed around explicitly permitted capabilities.

  1. 01

    Owner policy

    Owners define what is remotely exposed. Capabilities are not public by default.

  2. 02

    Local safety boundary

    The architecture keeps safety-critical authority local. Cloud policy may be stricter; it cannot weaken local limits.

  3. 03

    Controlled exposure

    Only permitted capabilities leave the site. Hardware does not become an unrestricted Internet device.

Designed on zero-trust principles for secure remote Hardware-as-a-Service.

Architecture and in-development controls — not a production security certification, and not a claim that physical safety hardware is already shipping.

What is real today

Versioned contracts and working SDKs — not a mockup.

The interactive experience on this site runs on simulated hardware, and says so. The engineering underneath it is real, versioned, and checked into a repository. These are separate facts, so they are stated separately.

REAL AND VERSIONED

  • TwinLink contracts, pinned to a specific commit with a SHA-256 checksum for every file
  • OpenAPI specification — 11 paths, 12 operations, under a /v1 prefix
  • JSON Schemas for resource, session, command envelope, receipt, acknowledgement, lifecycle, telemetry frame, audit event and errors
  • Canonical device-model capability catalog — telemetry paths, command arguments, units, safety declarations
  • TwinLink Client SDK (TypeScript and Python) and Connector SDK (Python)
  • Conformance kit and Bridge runtime
  • Access, lease, grant and local-safety semantics — 193 passing tests in this repository

SIMULATED IN THIS EXPERIENCE

  • Hardware execution — served by a simulated provider, never a physical device
  • Telemetry frames and command results, all marked with simulated provenance
  • Owner, mission, laboratory and manufacturer personas
  • Cross-organization grants — an experience-layer capability, not yet a TwinLink API

NOT CONNECTED YET

  • Physical spacecraft hardware
  • Production identity — OIDC and mutual TLS are designed, not implemented
  • A physical Safety MCU; the safety interface is architecture-frozen, not built
  • Certification of any kind — no security, safety, space or flight qualification is claimed

The workflow does not change when the hardware connects.

Hardware execution is simulated today. The product is built so a future TwinLink API provider implements the same interface against physical hardware — the same resources, the same leases, the same commands, the same evidence. A conformance stub for that provider already exists in this codebase and deliberately refuses to run, so the boundary cannot be blurred by accident.

Two versions, stated separately

This website and the TwinLink repository move at different speeds. Conflating them would overstate what the site actually runs.

This website runs on
TwinLink contracts 0.3.0
Pinned to commit 3332261 with a per-file checksum: 11 JSON schemas, 1 OpenAPI specification, 1 capability catalog. Unmerged fault-injection contracts are deliberately not used here.
The TwinLink repository is at
Contracts 0.4.0 (frozen) · SDK 0.3.1
21 JSON schemas and 2 OpenAPI specifications, including a frozen fault-capability surface. Those contracts are frozen; runtime fault execution on physical hardware is not complete and is not claimed.

Implementation path

From architecture to connected hardware.

The website deliberately separates what you can interact with today from measured results that appear only once a physical resource is connected.

  1. 01

    IN DEVELOPMENT

    TwinLink interoperability platform

    The common device, access, orchestration, and connector layer that the rest of the experience is built around.

  2. 02

    INTERACTIVE · SIMULATED

    Role-based product experience

    Mission teams, manufacturers, laboratories, and strategic partners can each run the full product workflow end-to-end, with the status of every result labelled.

  3. 03

    NEXT PHYSICAL PROOF

    CubeSTEM reference hardware

    A controlled CubeSTEM resource can move the experience from simulated telemetry to measured hardware results through the same TwinLink path.

  4. 04

    PARTNER-DEPENDENT

    Founding third-party integrations

    External hardware becomes a real network resource only after technical integration, owner approval, and an agreed access policy.

Founding programme

Two ways to be among the first.

Hardware owners come from two directions: manufacturers who want their products evaluated before procurement, and laboratories with capable equipment that is idle most of the week. Both keep control of what is exposed, to whom, and for how long.

Founding hardware partner programme

Make your hardware testable before it ships.

Selected founding spacecraft-hardware manufacturers may receive an initial TwinLink integration at no integration charge, subject to technical fit and mutually agreed scope.

The manufacturer remains in control of what is exposed, who may access it, and what commercial relationship follows an evaluation.

We are looking for

If you build an OBC, EPS, ADCS subsystem, payload, sensor, SDR, actuator, or other spacecraft engineering hardware, we would like to explore a founding integration.

Technical fit + mutually agreed scope

Founding laboratory node

Make one resource reachable without giving up your lab.

Selected founding laboratories may receive an initial TwinLink integration for one resource at no integration charge, subject to technical fit and mutually agreed scope.

The laboratory decides what is exposed, who may reach it, and when. Private first; shared only if and when you choose. Nothing is installed beyond software you run yourself, and no equipment leaves your site.

We are looking for

If you run a cubesat programme, a FlatSat bench, an ADCS or reaction-wheel table, thermal or vacuum test equipment, a ground station, or comparable laboratory hardware, we would like to explore a founding laboratory node.

No appliance to install · no inbound port