Jaeger Previews AI-Assisted Troubleshooting and GenAI Trace Visualization

SALT LAKE CITY -- Jaeger, the CNCF-graduated distributed tracing project, is previewing its first integrated AI-assisted troubleshooting and generative AI (GenAI) observability capabilities at KubeCon + CloudNativeCon North America 2026. The experimental features are available in Jaeger v2.22.

Ask Jaeger and Native MCP Tools

The preview introduces Ask Jaeger, an in-product natural-language assistant that lets engineers investigate traces conversationally, with visible tool calls that show how the assistant reaches its conclusions. An opt-in Model Context Protocol (MCP) endpoint provides nine tools for discovering services and traces, examining errors, mapping service topology, identifying critical paths and inspecting individual spans. Built-in diagnostic skills for error root-cause analysis and N+1 query detection give AI agents repeatable investigation workflows.

The maintainers said a central design principle is progressive disclosure, in which AI agents retrieve only the trace evidence needed for each investigative step rather than sending an entire trace to a language model. The architecture combines MCP, Agent Client Protocol (ACP), AG-UI and OpenTelemetry, and works with any ACP-compatible agent rather than a single model or proprietary framework.

GenAI-Aware Trace Visualization

New specialized views based on OpenTelemetry GenAI semantic conventions help teams examine model interactions, agent activity, retrieval operations, token usage, system instructions and tool execution within distributed traces. Developers can enable the experimental capabilities after downloading Jaeger and following the project’s AI sidecar integration guide.