GQueries: Grounded memory and authorized evidence for production AI

We built GQueries because retrieval alone is not enough for production AI. Finding a document, memory, or passage does not automatically mean an AI system should be allowed to use it in an answer. GQueries adds a controlled layer between retrieval and generation: persistent memory, authorization, provenance, evidence selection, and fail-closed grounding. The goal is to help applications answer a few important questions before delivering a response: Is this evidence relevant? Is it authorized for this requester? Is the claim actually supported? Should the system answer, or abstain? GQueries is designed to work with existing LLMs and retrieval stacks through BYOK, rather than replacing them. We’re launching it as infrastructure for teams building agents, support systems, enterprise assistants, and other AI applications where unsupported or unauthorized answers are not acceptable. I’d love to hear how you currently handle grounding and evidence validation in production.