Limit Platform is the operating system for sovereign AI.
Apps on top. Sovereign infrastructure underneath. Every layer runs inside the customer perimeter, on customer hardware, under customer control.
// apps on top · sovereign infrastructure underneath · nothing leaves the perimeter
Four layers. Apps on top. Sovereign infrastructure underneath.
Limit Platform is structured the way operating systems are. The foundation is fixed. The applications above are replaceable.
Inside the perimeter. End to end. On your hardware.
Limit Platform deploys on customer infrastructure. Single-site, multi-site, or fully air-gapped. The platform fits where the work is and where the regulator says it has to stay.
The customer is the controller of every byte the platform processes. Your data stays inside the perimeter. Your model weights stay. Your prompts stay. Your evidence trail stays.
Modularity is a deliberate design choice. Contract-first interfaces between layers mean a customer can swap a component, replace a vendor, or take the deployment in-house. No lock-in disguised as integration.
The hard 80%, off the shelf.
Running a model is the easy 20% of any production AI system. Governance is the other 80%. Identity. Authorization. Audit trail. Evidence retention. Policy enforcement. Sensitive-data handling. Customer-system integration. The work that turns a model into something the auditor, the CISO, and the CFO can sign off on.
Limit Platform ships this layer out of the box. Identity federates with the directory systems already inside the perimeter. Every action, prompt, and inference lands in the audit trail. Authorization is policy-driven and enforced at every layer. Sensitive data is discovered, classified, and governed before AI touches it.
Every layer you need, end to end.
Each capability is a layer of the operating system, built once and shared across every application. Open any one to see how it works and how it deploys inside your perimeter.
How the operating system is built and where it runs.
The controls the auditor, the CISO, and the regulator rely on.
- Limit Trace
Audit and evidence trail
Every action, every prompt, every inference in the audit trail.
- Limit Control
Authorization and policy
Policy decides what AI can do. Enforced at every layer.
- Limit Core · Limit Vault
Identity and access
Federated with the directory systems already inside the perimeter.
- Limit Data
Sensitive-data handling
Sensitive data is governed before AI sees it.
The layers that move, store, and execute the work.
- Limit Store
Storage and artifacts
Versioned, content-addressed storage for everything the platform needs.
- Limit Bus
Messaging and events
Durable event streams that don't leave the perimeter.
- Limit Runtime · Limit Deploy
Compute and jobs
AI workloads run on customer compute.
- Limit Runtime
Models and inference
Bring your own models. Inference stays inside.
- Limit Bus · Limit Core
Customer-system integration
Connect to the systems already inside the perimeter.
See Limit Platform in your environment.
Whether you're sizing it up for a single line, a fleet, or an entire perimeter, the next step is a conversation.