About us
The idea to build a Kernel for AI agents was born from the pain of building agents that deal with complex real world data.
We began as an insuretech startup. Our first goal was simple: Extract information out of insurance documents. It was supposed to be the core around which we built every next feature, building progressively more complex agents and workflows.
In reality, however, we had massively underestimated all parts of the task. Not only was extracting the data much more difficult than we had assumed, connecting different flows and agents which each interacted with the data under a different lens created a convoluted patchwork of eccentric features. We found that even maintaining our features was a massive chore, while connecting features that should have been connected often was too difficult to be worth the effort.
We truly missed the luxury of all the tooling, scaffolding and systems available when building outside of LLMs. So we decided to build it.
