Orin
An agentic harness built specifically for smaller local models, able to write code and run scheduled tasks. In development, not yet public.
Most agent frameworks assume you are calling a large hosted model. Orin starts from the opposite premise: that a smaller model running on your own hardware can do real, useful work if the harness around it is built well.
Orin is that harness. It is designed for smaller local models, and it can write code and run scheduled tasks on its own. The interesting work is in the scaffolding, not the model: giving a modest model enough structure, memory, and guardrails to be reliable without leaning on a giant model to cover the gaps.
It is still in development and not publicly available yet. I am building it because the ability to run capable agents locally, without sending your work to someone else’s API, is going to matter more every year.