Product

A compute capacity abstraction for modern workloads

NoirLedger keeps the customer experience centered on capacity, uptime and execution while the control plane handles placement.

Principle

The product is compute capacity

The product remains pool-based, API-first and audit-friendly, with no raw node selection in the default UX.

Principle

Customers request resources, not machines

The product remains pool-based, API-first and audit-friendly, with no raw node selection in the default UX.

Principle

Protected and best-effort lanes

Best-effort is optimized for low-friction experimentation. Protected adds reserved failover capacity and minimum replica policy.

Principle

One workload abstraction

The same workload object covers long-running services and finite execution without turning the product into a VPS or job marketplace.

Principle

Wallet-native billing and network economics

NLC handles product billing. NLR stays available for worker rewards and future staking mechanics.

Principle

Hybrid supply model

The product remains pool-based, API-first and audit-friendly, with no raw node selection in the default UX.

Principle

API-first and SDK-first usage

The product remains pool-based, API-first and audit-friendly, with no raw node selection in the default UX.

Principle

Security, isolation and observability

The product remains pool-based, API-first and audit-friendly, with no raw node selection in the default UX.

Service workloads

Run APIs, bots, services and backend systems as long-running workloads with optional protected uptime.

Finite workloads

Run jobs, pipelines and burst inference without keeping capacity online when nothing is running.

SDK and API

Request capacity, deploy workloads, stream logs and manage usage with simple primitives.

Worker supply

Turn controlled local hardware into network capacity and choose how much of your machine you share.

Architecture

Eight layers keep the product coherent end to end

The customer sees one compute fabric, while the platform stays split into clear operational layers.

Client surfaces

Marketing site, docs, dashboard, onboarding, billing UI and worker console.

Control plane core

API ingress, auth, wallet, ledger, billing, scheduling, routing, logs and workload state.

Compute pool

Trusted core capacity and community workers contribute CPU, RAM, GPU, VRAM and network into one pooled fabric.

Worker runtime

A Noir worker agent registers hardware, receives jobs, launches microVMs and reports telemetry under KVM-backed policy.

Fabric policy

Quotes, reserve logic, placement intent, failover rules, observability and payout rails stay under the control plane.

Security and access

Operational trust sits inside the product model

Identity, isolation, ledger correctness and worker trust controls are treated as product requirements.

Authentication baseline

Email verification, password reset, strong sessions and TOTP 2FA are part of the operating model.

Secrets and execution isolation

Secret visibility stays minimal and workload execution uses restricted, auditable isolation boundaries.

Financial correctness

Ledger entries, usage records and top-up reconciliation are designed to be auditable and replay-safe.

Worker trust controls

Signed heartbeat, benchmark enrollment, validation tasks and quarantine are treated as first-class.