How Jaeger is evolving to trace AI agents with OpenTelemetry

Jaeger is evolving in two phases to handle AI workloads. First, Jaeger v2 rebuilt its core around the OpenTelemetry Collector, natively ingesting OTLP to unify metrics, logs, and traces. Second, the project is integrating Model Context Protocol (MCP), Agent Client Protocol (ACP), and Agent–User Interaction Protocol (AG-UI) to enable collaboration between engineers and AI agents during incident debugging. A new backend ACP layer translates natural-language queries into deterministic trace queries, supporting both cloud LLMs and local SLMs for data privacy. The UI is being updated with a streaming assistant powered by AG-UI. Additionally, Jaeger is adding visualization support for OpenTelemetry GenAI semantic conventions, covering RAG pipelines, agentic systems, embedding latency, and token usage tracking.