No training by default
We do not use your data or memory to train models by default. Any future change to that position should be explicit and opt-in.
Privacy and principles
Control should be a product feature.
These are the commitments guiding WulffAgent as it takes shape. They describe the direction honestly, including what is still an intention rather than a finished capability.
Join early accessYou own your data
Your personal knowledge, memory, and agent configuration belong to you. WulffTech Account is the identity root, not a claim on the things you create.
Export and deletion
You should be able to export your data in a useful format and request deletion. Those controls are part of the product direction from the beginning.
Local-first options
We intend to support local-first and hybrid ways to use the system, alongside zero-knowledge options where they make sense.
Permission before action
Agents should work within a scope you explicitly set. Actions should be understandable before they happen and recorded in an audit trail afterward.
WulffTech Cloud remains available as a secure data layer today. Its availability does not change these principles for the personal AI products we are building.