MemoryGate

MemoryGate is a pre-LLM memory gating layer that evaluates retrieved memories from vector stores and returns relevance, trust, and confidence scores. It solves the problem of outdated or conflicting information in AI knowledge bases by providing signals rather than enforcing hard policies. It is designed for enterprise teams in HR, legal, compliance, and onboarding, as well as internal search. It is positioned as a signal-only layer that gives downstream consumers control without data loss, and it is delivered as an API.

Key features

• Returns relevance, trust, and confidence scores
• Surgical Vector Trust Decay (SVTD)
• Privacy mode with zero content storage
• No hard policy enforcement
• Signals for downstream filtering