Three years ago, I was asking myself a question: what problems can we really solve with generative AI? I quickly realized that getting a model to produce an answer was only a small part of the work. Building useful applications required solid foundations, and governance had to be part of them: deciding which models run, what data they can access, and what actions they can take. I wanted to benefit from this technology without becoming dependent on a provider or giving up control of our data. That became the starting point for LM-Kit. My motivation was to make generative AI affordable and accessible to companies and individuals who care about privacy and sovereignty. Adopting it should not require handing over your data or having a large engineering team build everything from scratch. Today, we’re launching LM-Kit One, a private AI application server that brings model inference, knowledge search, agents, and orchestration together through a REST API. It runs on your own infrastructure, with the models you choose. To support that goal, it’s free for personal and educational use, and for small companies that meet our EULA’s eligibility criteria, including production use. Evaluation and development are free for everyone, regardless of company size, with no time limit. There’s still plenty of engineering ahead, but I’m happy to share what we’ve built. Making this technology useful to more people around the world, while keeping it under their control, is something I’m glad to contribute to.

LM-Kit One

The Private AI Application Server

The Private AI Application Server