Standalone proof
Use the app's embedded engine and default model to prove one bounded workload on real endpoint hardware.
Private AI for organisationsYour models.
Your models.
Your infrastructure.
Your operating rules.
Every AI App can begin independently with its embedded AI Engine. Add AI Server when models, compute or APIs need to be shared; add AI Gateway only when several servers need one resilient front door.
Optional scale-out architecture
Standalone first. Add each shared layer for a reason.
AI Apps retain their embedded engine and local route even when they connect to the organisation platform shown below.
AI Apps · embedded engine retainedYour applicationsOpenAI-compatible tools
Optional encrypted service path
AI Gateway · multi-server onlyDiscovery, health-aware routing and failover
AI Server 01chat · embeddings
AI Server 02image · vision
AI Server 03speech · capacity
AI Admin Console · when governedIdentity · licence · policy · audit · usage
Adoption path
Buy the control you need when you need it.
Shared AI Server
Add one central model, compute or API service only when the business case needs it.
Managed estate
Add members, entitlements, server enrolment, policy and audit evidence when central governance matters.
Gateway pool
Put AI Gateway in front of several server workers when resilience, specialisation or measured capacity requires them.
Operational fit
Designed for infrastructure teams, not only AI demos.
Deployment options include Windows, macOS, Linux, Docker and Kubernetes. The right topology depends on workload, hardware, identity, availability and compliance needs.
Familiar APIs
Connect compatible clients to chat, embedding, image, speech and vision services.
Chosen boundary
Place models and inference on infrastructure controlled by your organisation.
Central administration
Manage members, licences, enrolment, policy, audit and usage evidence.
A path to scale
Move from one node to a gateway-managed farm as demand and evidence grow.
Turn requirements into a decision
Open deployment planner Find the smallest architecture that fits.
The deployment planner evaluates embedded apps, optional servers, Gateway, administration, hardware and paid licence quantities—then sends only orderable lines into the Store.
14current-generation products available
1M+app-family downloads each month
6+Fortune Global 500 customers
0project failures since 2007
Bring one real workload
Talk to an enterprise specialist