Nacre - AI Context Layer

Nacre is a self-hosted knowledge index designed for agent systems, where access permissions are integrated directly into the retrieval process. Key features: * Permission-aware index traversal * Self-hosted for data residency * Connects via MCP or REST * Docker deployment in minutes * Ensures users and agents see only authorized content This system fundamentally changes how information is accessed by agents. Unlike traditional methods that filter results after retrieval, Nacre executes access checks during the index traversal. This guarantees that `top_k` results always represent `k` permitted documents, never silently stripping information afterwards. It means if a document is unauthorized, it's as if it doesn't exist to the querying agent or user, providing a robust security posture. Built for organizations requiring stringent data access controls and privacy, Nacre ensures that every document remains on your own machine. It's an ideal choice for enterprises managing sensitive information, allowing them to maintain full control over their knowledge base while enabling agent systems to operate within defined security perimeters. The open-source core is Apache 2.0 licensed, with commercial modules available for advanced functionalities like multi-tenancy, SSO, SCIM, deny rules, and SIEM export. Perfect for engineering teams, security officers, and compliance departments who need to ensure data access integrity within their internal systems and for agents operating on behalf of individuals.