Architecture that runs end to end inside the perimeter.
Limit Platform is structured as a four-layer operating system, deployed on customer infrastructure. Every layer runs where the data and the work live, with contract-first interfaces between layers.
// architecture · inside the perimeter
Four layers. Sovereign at every layer.
At the bottom: Platform Core, the sovereign primitives every application uses. Identity, storage, audit, messaging, deployment, secrets, vault. Above that: Platform Suites, the execution layer for AI workloads. Limit Runtime executes inference, jobs, and agents; Limit Data prepares and governs data.
Above the platform suites: Product Suites, the business-capability layer. Limit Standard, Limit Signal, Limit Docs, Limit Knowledge. On top: Vertical Solutions, the industry-packaged applications customers actually deploy. Each vertical inherits the sovereignty, governance, and modularity of every layer below.
A deployable architecture, not a reference model.
- 01
Sovereign foundation at the bottom
Platform Core provides the primitives every application uses: identity, storage, audit, messaging, deployment, secrets, vault. Built once, shared across the platform, deployed inside the customer perimeter.
- 02
AI runtime as a platform layer
Platform Suites (Limit Runtime, Limit Data) execute AI workloads and prepare data, both on customer compute. The runtime is model-agnostic; the data layer is policy-aware.
- 03
Business capability and vertical packaging on top
Product Suites and Vertical Solutions give industries the shape they buy: workflows, data models, evidence schemas, and integrations matched to the vertical, all built on the same foundation.
- 04
Contract-first interfaces between every layer
Each layer exposes a documented, versioned interface. Customers can swap components, replace vendors, or take a deployment in-house without rewriting the operating layer.
Deploys where the work is. Single site, multi-site, air-gapped.
Limit Platform deploys on customer-provided hardware in the environment the work lives in. Single-site deployments fit a plant, a firm, or a single SOC. Multi-site deployments coordinate across plants or vessels with consistent governance. Air-gapped deployments operate without any external network dependency at runtime.
No off-perimeter dependency for inference, governance, or evidence. Aggregated metrics may sync to a customer-controlled aggregation point if the customer DPA permits it. No telemetry to Limit Systems by default.
- Hardware
- x86 Linux, virtualized or containerized. GPU acceleration where the workload requires it. Customer procures or we pass through at cost.
- Operating environment
- Customer-managed Kubernetes, VMs, or bare metal. Multi-tenant inside a single customer perimeter; cross-customer isolation by deployment, not by tenant boundary.
- Network posture
- No outbound calls required for runtime operation. Aggregation sync (if configured) is customer-initiated.
- Failure modes
- Designed for intermittent and absent connectivity. Air-gapped operation is a first-class deployment mode, not an afterthought.
This capability is one layer of the operating system. Every application uses it. Every customer perimeter that runs Limit Platform gets it. See the full architecture →
See it in your environment.
Walk us through the systems already in place. We'll show you how this layer fits.