I built GovernsAI to solve a problem I kept hitting while switching between OpenAI, Anthropic, and Google: no shared memory, no centralized access control, and PII leaking into prompts constantly.

It's essentially an AI OS layer that sits above the providers:

- Unified authentication across OpenAI, Anthropic, Google - Persistent memory management that follows you across models - A precheck service that catches PII before it hits any API - Budget enforcement and human-in-the-loop confirmation workflows - A browser extension (pii-guard) that intercepts at the input level

The architecture is documented in a paper I submitted to arXiv if you want to go deep on the design decisions.

Happy to answer questions about the infra choices, the memory layer, or why I built on top of providers instead of picking one.

https://governsai.com

Github: https://github.com/Governs-AI