Defence engineering
Develop and maintain software inside restricted networks and operational facilities.
Private AI coding infrastructure Defence and critical industries
Your AI coding agent. Anywhere. Even completely offline.
Bring the capabilities of AI-assisted software development to air-gapped, restricted and forward-deployed engineering environments.
A dedicated Mac mini runs an open-weight coding LLM locally, while engineers work from their own laptops using familiar development tools.
- Runs an open-weight coding model on a dedicated Mac mini
- Engineers work from their own laptop
- No internet. No cloud API. No project data intentionally stored on the AI server.
Designed for environments where the cloud is not an option
Forward deployment
Bring an AI coding assistant to remote sites, field deployments and mobile engineering teams.
Critical infrastructure
Support industrial automation, energy, transportation and other restricted operational environments.
Sensitive development
Keep model inference local without transferring proprietary source code to external AI providers.
Architecture
How it works
The engineer's laptop holds the project files and runs the coding agent. The Mac mini provides local AI inference through a dedicated Thunderbolt network connection. Code changes and tests execute on the laptop.
Engineer's laptop
Project files, the coding agent and tests. VS Code, or a compatible editor.
Mac mini
Headless. Ollama serves an open-weight coding model. No project store by design.
-
01
Connect
The engineer connects a MacBook to the headless Mac mini using a Thunderbolt USB-C cable.
-
02
Request
The AI coding agent sends model requests directly to the Mac mini. The Mac mini hosts the open-weight coding model using Ollama and returns responses.
-
03
Edit and test
The agent can inspect the project, propose modifications, edit files and execute tests with the engineer's approval. That work stays on the laptop.
-
04
Stay offline
After initial provisioning, the system operates without an internet connection.
Controls
Security by architecture
- No cloud inference or external AI API dependency
- Source code remains on the engineer's laptop
- No persistent project or conversation storage on the Mac mini by design
- Headless operation with preconfigured connectivity
Commercial offer
Simple deployment. Predictable pricing.
The entry package is one AI server and one engineering workstation. That boundary is fixed so a larger rollout is quoted as a larger system, not absorbed into this price.
One-time system deployment
$5,000 USD
per setup
- Initial installation and model deployment
- Agent configuration and direct laptop connectivity
- Offline validation
- Technical documentation and engineer onboarding
Annual software and support licence
$2,000 USD
per year
- Maintained configuration
- Compatible software updates
- Support
- Offline update procedures under agreed service terms
Included in this price
One AI server and one engineer. Hardware is purchased separately.
Quoted separately
Additional users, further AI servers, bespoke security hardening, accreditation activities and on-site travel.
Demonstration
Bring AI-assisted engineering into restricted environments
Enable engineers to build, debug and maintain software without relying on external connectivity or cloud-hosted AI services.
- Private.
- Portable.
- Offline by design.