The product
What xAIDR is building
xAIDR is an open-source runtime-security component for teams building AI agents. Installed as a Python package, it sits on the agent's execution path to inspect inputs, tool calls, model outputs, and agent-to-agent messages before an action proceeds. The project documents local policy controls, approval gates, structured telemetry, and destinations including stdout, files, webhooks, and OpenTelemetry. Its documentation emphasizes an in-process scan path without a required backend or account, giving developers a practical way to add inspection and policy enforcement to agent workflows.
From the maker
A fresh open-source control point for agent actions
xAIDR is relevant to founders experimenting with agentic workflows because it focuses on the operational boundary where an agent can call tools or hand work to another agent. The publicly documented install path and local policy model make it straightforward to evaluate without treating claimed performance or enterprise features as established outcomes.