Introduction

Welcome to the documentation for ZeroTrust Agents.

ZeroTrust Agents is an open-source, LLM-agnostic gateway proxy that implements Policy-as-Code for AI agents. It acts as a semantic firewall between your AI agents and the LLMs they use to execute tools.

Core Capabilities

• 🛡️ Semantic DLP: Uses smaller local LLMs to inspect tool payloads for PII, leaks, or malicious intent before executing them.
• 👤 Human-in-the-Loop: Suspend agent execution and ping a human on Slack/Teams or via the API to approve high-risk actions.
• 🤖 LLM Agnostic: Drop-in proxy support for OpenAI, Anthropic, Gemini, Ollama, and Model Context Protocol (MCP) compatible tools.
• 📝 Policy-as-Code: Define agent tool access limits and security policies in clean, declarative YAML.