Tracing AI Agent Execution: Foundations, State of the Art, and Implications for Distributed Multi-Agent Systems

As of October 2026, the major AI-agent platforms can record the individual steps of an execution, yet theecosystem still lacks stable and uniform semantics. This paper combines three contributions: a conceptualframework for agent-observability questions; a comparative review of 13 platforms, with package versionschecked on 10 October 2026; and a non-normative interpretation of the implications for SGAEIA in distributedmulti-agent and edge environments. The results show that a trace reconstructs the path of execution but doesnot demonstrate the correctness of the outcome; that essential capabilities remain in beta, preview, orDevelopment status; and that divergent content-capture defaults create direct privacy, governance, andevidence-continuity risks. The central conclusion is that tracing, evaluation, and governed evidence mustoperate as complementary functions: observing execution, judging outcomes, and preserving a verifiable basisfor audit and assurance.