Welcome to OptimaGPT

OptimaGPT is an on-premises AI server built by Parker Software. It gives your organisation access to powerful large language models without sending any data to a third-party service. Everything runs on your own hardware, on your own network.

Why on-premises?

Cloud AI services are convenient, but they come with trade-offs. Every message sent to a cloud provider leaves your network. For organisations handling sensitive data — whether for compliance, confidentiality, or data sovereignty reasons — that is not acceptable.

OptimaGPT is designed for exactly these situations. Your data stays with you. Models run locally. Nothing is transmitted externally.

OptimaGPT exposes an OpenAI-compatible API, so any application already built to work with OpenAI can be pointed at OptimaGPT instead with no code changes required.

The three components

OptimaGPT consists of three components that work together:

OptimaGateway is the central hub. It handles authentication, request routing, and the web interface. Administrators configure everything — users, API keys, connected nodes, and AI models — through the Gateway's admin panel. The chat interface is also built in to the Gateway.

OptimaNode is the inference engine. It discovers AI model files on its host machine, manages the underlying inference processes, and executes requests forwarded to it by the Gateway. Nodes can run on the same machine as the Gateway or on separate hardware, and you can connect multiple Nodes to a single Gateway.

OptimaChat is the chat interface. Accessible from any web browser by navigating to the Gateway's address, it lets users have conversations with your AI models. The same interface is also available as the OptimaGPT desktop application.

Who uses what

Role	Component
IT administrator	Gateway — configuration, users, nodes, model management
Inference machine	Node — hosting model files, executing requests
End user	OptimaChat — browser or desktop app
Developer / integrator	Gateway API