GovernsAI – unified auth, memory, and PII guard across AI providers

Name: GovernsAI – unified auth, memory, and PII guard across AI providers
Availability: InStock
Author: shaivpidadi

by shaivpidadi·Mar 5, 2026·2 points·0 comments

Visit Project View on HN

AI Analysis

●●●BangerBig BrainSolve My ProblemZero to One

PII-blocking precheck + centralized memory across any AI provider beats building this yourself.

Strengths

•Genuine architectural insight: precheck interception layer (PRESIDIO + FastAPI) catches PII before API submission—solves a real compliance gap
•Persistent memory + OIDC across providers addresses actual multi-agent fragmentation (no other tool unifies memory this way)
•Provider-agnostic design (sidecar SDK, proxy mode) means zero lock-in—customers control keys, policies, and audit trails

Weaknesses

•Enterpise GTM required; pilot program suggests early stage—adoption risk and pricing still unknown
•Browser extension pii-guard could leak elsewhere; full coverage requires app redesign or extensive agent rewrite

Post Description

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