Xaidr Provides In-Process Runtime Security for AI Agents
TL;DR. Delphisecurity released Xaidr, an open-source tool for in-process runtime security and governance of AI agents. - Xaidr directly inspects agent actions and messages to block prompt injection and destructive tool calls. - The tool operates locally without external dependencies, ensuring data privacy and security. - It offers both monitoring and enforcement modes for flexible deployment within agent systems.
- Xaidr is an open-source, in-process runtime security tool for AI agents.
- It inspects agent inputs, tool calls, model outputs, and agent-to-agent communication.
- The tool prevents prompt injection, jailbreaks, data leakage, and destructive actions locally.
- Xaidr functions without a backend, account, API key, or network in its core scan path.