Runtime security guardrails for AI agents — inspect, control, and audit every tool call. Zero-dependency local mode included.
agentguardx 0.1.0
The release of agentguardx 0.1.0 introduces runtime security guardrails for AI agents, enabling inspection, control, and auditing of tool calls with a zero-dependency local mode. This tool is designed to mitigate risks in AI agent ecosystems but could introduce unintended vulnerabilities if misconfigured or exploited by adversaries targeting AI workflows. Users deploying AI agents must assess its security implications to prevent unauthorized tool call manipulations or data exfiltration.