In your existing environment
We deploy into your premises, network and security model, integrating with the systems your team already trusts.
Private AI. Fully supported.
Use AI on sensitive work. Keep it under your control.
We design, supply, deploy and support a complete private AI stack—in your premises, on hardware sourced by us, or on dedicated cloud compute.
One accountable partner for infrastructure, models, controls, integration and ongoing support.
Your private AI environment
A clear boundary, operated as one stack
Your protected data boundary
Business workflows and agents
Maix controls and integrations
Private models and inference
Dedicated compute
On premisesSupplied hardwarePrivate cloud
Built for organisations that need useful AI without an unexplained data journey
- Clear data boundary
- Private model runtime
- Human approvals
- Searchable audit
- Ongoing support
Your environment, your boundary
Private AI deployed where it makes sense for you.
Start with the security and operational reality of the business. We choose the deployment model around that—not the other way round.
On hardware sourced by us
We specify, source and configure the right GPU hardware, then install it as a supported private AI environment.
On dedicated cloud compute
Run in an isolated cloud environment selected for your workload, data boundary and preferred region.
The complete stack
Not another tool for your team to assemble.
Private AI fails when every layer has a different owner. We bring the infrastructure, software and operational work together as one supported environment.
You keep ownership of the boundary and the business decisions. We take responsibility for making the technology work inside it.
Compute and infrastructure
Sizing, hardware selection, networking and resilient deployment matched to the work you need to run.
Private models and inference
Models selected, tested and served inside your chosen boundary rather than assumed to be a public API call.
The Maix control layer
One governed workspace for models, agents, tools, tasks, memory and browser sessions.
Business integrations
Connect the systems people actually use, with access scoped to the right workspace and purpose.
Security and governance
Permissions, human approvals, protected credentials and searchable evidence for sensitive actions.
Ongoing human support
Updates, monitoring, recovery and a team that remains accountable after the first deployment.
Private by deployment, not by promise
Controls people can understand and verify.
Security-conscious teams should not have to accept a vague claim that their data is safe. Maix makes the boundary, permissions and evidence part of the working system.
A boundary you can explain
Choose where models run, which systems they can reach and whether any workload may leave your environment.
Credentials stay protected
Connections are scoped and secrets are encrypted, injected only for the capability that needs them.
People keep the final say
Sensitive or destructive work can stop for approval before an agent is allowed to continue.
Every action leaves evidence
Tool use, decisions, approvals and hand-offs remain attributable and searchable without recording secret values.
Technical detail
Maix supports local model serving, workspace-scoped tool access, encrypted credential storage, approval gates, command and tool restrictions, sandboxing, searchable audit history and optional connections to explicitly approved external services.
Fully supported
From the first boundary diagram to everyday use.
A private AI deployment is an operating capability, not a box delivered at the door. We stay accountable as the environment and the work evolve.
Talk through your environment- 01
Define the boundary
We map the sensitive work, systems, users and security constraints before choosing hardware or models.
- 02
Build the right stack
We provision compute, qualify the models, deploy Maix and integrate the tools that create useful outcomes.
- 03
Put it into real work
We start with bounded use cases, train the people responsible and prove the controls with observable evidence.
- 04
Support it properly
We monitor, update and improve the environment while keeping rollback, recovery and ownership clear.
Sensitive work is still useful work
Bring AI to the information you cannot casually upload.
We begin with a valuable, bounded workflow and build outward only when the organisation can see that the controls hold.
- Client and legal material
- Commercial and financial analysis
- Health and care information
- Product designs and intellectual property
- Internal operations and communications
- Regulated or contract-bound data
The technology underneath
Open where trust benefits from inspection.
Maix is the open-source control layer behind the deployment. It keeps models, agents, tools, memory, approvals and activity consistent across the AI clients your team chooses.
Inspect the sourceOne governed working layer
Across local models, agents and business tools
Claude CodeCodexOpenCodeCursorGrokGemini CLIPrivate modelsCustom agents
Find out what private AI should look like in your environment.
Start with the sensitive work, the boundary it needs and the people responsible. We will turn that into a practical deployment plan.