DocsGetting Started

Getting Started

Introduction to vSAML

vSAML (Virtual Security Assertion Markup Language) is an architectural solution designed to address the security flaws inherent in RAG systems.

The Problem

Modern AI agents can:

  • Read sensitive files and immediately send them to external services
  • Execute arbitrary code without verification
  • Access databases and leak confidential information

    • The Solution

    vSAML introduces an **interceptor layer** between your AI agent and its tools. This interceptor:

    1. **Tracks data flow** using taint analysis

    2. **Enforces static rules** (ALLOW/DENY per tool)

    3. **Applies dynamic rules** based on tool sequences