Lab of AI builds a portfolio of agent-based applications for service operations, analytics, planning, procurement, supply chain, and decision support — all powered by Kyphren, our shared runtime for reliable multi-agent systems. Kyphren runs on any major public cloud and across model providers, so deployment follows the customer’s infrastructure rather than dictating it.
What we're building
Enterprises do not need another agent framework. They need AI that performs real work inside existing processes — answering service requests, interrogating data, planning demand, supporting decisions — without replacing the systems they already run on.
We build those applications as products, not as bespoke projects. Each one is designed to be deployed and configured for a customer, not rebuilt from scratch for them.
Kyphren is what makes that possible. It provides the agent coordination, capability discovery, durable execution, persistent state, memory integration, and observability that every one of these applications would otherwise reimplement — across different model providers and model types, including large language models, predictive machine learning, and computer vision.
Architecture
Our applications sit on top of a shared runtime. Kyphren separates them from agent coordination, model providers, and infrastructure services — so a capability built once is available to all of them.
Applications
Kyphren Runtime
Providers & infrastructure
Model providers
Hosting
Enterprise integrations
Reference application
Our AI-powered service desk is the furthest along: it already handles multi-agent coordination, conversation management, persistent state, external knowledge retrieval, authentication, and operational logging inside a concrete business workflow.
It is the proof that the approach holds. Every capability it exercises goes back into Kyphren, so the applications that follow start from a working foundation instead of a blank page.
Who it's for
Our customers are organizations that want AI doing measurable operational work — not a chatbot bolted onto a website, and not a toolkit their team has to assemble. They get working applications that integrate with the systems they already run, with governance, cost control, and audit built in rather than added later.
The application portfolio spans enterprise functions:
Applications will be offered as a managed service, combining licensing with usage-based components. Customers can also run them within their own cloud when integration, governance, or data-residency requirements demand it.
Current stage
A working AI service desk validates the architecture through a concrete operational workflow.
Coordination, capability discovery, persistent state, provider integration, memory, and observability are being consolidated into the Kyphren runtime.
Further applications — analytics, planning, procurement, and decision support — are in development on the same runtime.
Private access is planned for Q4 2026, ahead of a broader early-access rollout.
Behind the lab
Lab of AI is built by senior engineers with decades of experience in enterprise software, cloud architecture, predictive machine learning, and computer vision systems used in production environments.
That background shapes how we approach agent reliability: as an engineering and operational problem, not only a model-selection problem.
Get in touch
We are preparing private access for organizations that want AI doing real operational work inside their existing systems.