Show HN: Traccia – open-sourced AI agent runtime governance (github.com)

Traccia has launched an open-sourced AI agent runtime governance toolkit, providing developers with a Python SDK designed to observe, evaluate, and enforce policies on AI agents and large language model (LLM) applications in real time. Built on OpenTelemetry, Traccia enables features such as distributed tracing, cost tracking, and runtime policy enforcement. Its framework-agnostic design ensures compatibility with popular tools like OpenAI Agents, LangChain, and CrewAI, making it a versatile asset for developers working with multi-agent systems. The significance of Traccia lies in its ability to implement governance structures that enhance the safety and accountability of AI agents, particularly in production environments. Key technical features include automatic instrumentation, guardrail detection, and zero-configuration setup, enabling effortless integration into existing workflows. Developers can define runtime policies, such as spending limits and model restrictions, while ensuring compliance with regulations like the EU AI Act. With features such as token and cost tracking along with offline evaluation mechanisms, Traccia aims to provide a robust infrastructure for assessing AI performance and maintaining high standards of governance and transparency in AI applications.