OpenTelemetry
A cloud-native observability framework that provides unified standards for collecting and exporting metrics, logs, and distributed traces. It is widely used for monitoring distributed systems.
Core Facts
OpenTelemetry was launched by the CNCF in 2019 following the merger of the OpenTracing and OpenCensus projects
可观测性的三大支柱框架包括Metrics（指标）、Logs（日志）和Traces（链路追踪）
OpenTelemetry is the de facto observability standard in the cloud-native domain.
Timeline (last 90 days)
A common industry approach to tracing is based on OpenTelemetry's span model, mounting each tool call and model generation as child spans under the root request, thereby supporting drill-down analysis by elapsed time or error rate.
A common industry practice is to uniformly export the fields returned by the AI SDK into OpenTelemetry span attributes or dedicated metrics, then connect them to observability platforms such as Langfuse and Helicone to achieve cost attribution by trace dimension.
LangSmith and OpenTelemetry's LLM semantic conventions are practices in the direction of Agent observability.
There are already LLM tracing solutions based on OpenTelemetry semantic conventions in the industry, which can connect Agent execution to existing monitoring systems.
For teams with high compliance requirements or those wishing to self-host, tracing can be built in-house based on standards such as OpenTelemetry, exporting LangGraph's node execution events to existing monitoring systems.
OpAMP (Open Agent Management Protocol) is an open protocol in the OpenTelemetry ecosystem for remotely managing collector fleets, supporting configuration delivery, status reporting and health checks.
Langfuse is an open-source project that supports self-hosting and integrates through a framework-agnostic SDK and the OpenTelemetry protocol, without binding to a specific orchestration framework.
MCP tool design should follow production practices: reduce model inference round-trips per tool call, cache the tools list with a TTL, use OpenTelemetry tracing (with context placed in _meta), and never silently modify a published tool—instead release a v2 version and deprecate the old one.
OpenTelemetry's GenAI semantic conventions are driving the standardization of LLM tracing fields, enabling traces produced by different frameworks to be collected and queried with the same backend.
Microsoft Agent Framework Tracing is responsible for showing how a request flows through the system (including routing, handoffs, and tool calls), Microsoft Foundry provides observability for production AI workloads, and the whole solution is based on OpenTelemetry, so traces can flow directly into the team's existing observability stack.
40 more timeline events
All Facts (20)
OpenTelemetry was launched by the CNCF in 2019 following the merger of the OpenTracing and OpenCensus projects
90%Verified可观测性的三大支柱框架包括Metrics（指标）、Logs（日志）和Traces（链路追踪）
80%VerifiedOpenTelemetry is the de facto observability standard in the cloud-native domain.
80%VerifiedThe OpenTelemetry standard has been extended to support tracing of LLM calls.
80%VerifiedOpenTelemetry是由Cloud Native Computing Foundation（CNCF）托管的开源可观测性框架，统一了分布式追踪、指标和日志三大信号的采集标准
75%VerifiedThe concepts of Trace and Span originate from Google's Dapper paper and the OpenTelemetry standard.
70%VerifiedOpenTelemetry (OTel) is a vendor-neutral observability standard covering three types of signals—distributed tracing, metrics, and logs—and defines how data is collected, propagated, and exported.
65%VerifiedIn traditional microservice architectures, distributed tracing relies on standards such as OpenTelemetry to link cross-service call chains via TraceID.
65%VerifiedOpenObserve incorporates Agent and LLM calls into the OpenTelemetry framework, enabling developers to uniformly analyze LLM calls alongside logs, metrics, and traces from backend services and databases on a single platform.
65%VerifiedIn agent scenarios, a complete agent task should be treated as a single Trace, with each LLM call, tool call, and memory retrieval being a Span.
65%VerifiedPhoenix is open-sourced by Arize AI and focuses on experiment tracking and evaluation for LLM applications, complementing Arize's commercial ML monitoring platform.
65%VerifiedObservability tools such as OpenTelemetry and LangSmith are applying distributed tracing concepts to monitor the execution chains of AI agents.
65%VerifiedOpenTelemetry (OTel) is an open-source standard led by the CNCF Foundation and is being progressively introduced into LLM applications to standardize data collection formats.
65%VerifiedAn AI Agent session is essentially an independent runtime context that completes tool registration during the initialization phase and fixes the tool registry as a snapshot from startup time; active sessions cannot dynamically detect configuration changes.
65%VerifiedOTLP (OpenTelemetry Protocol) is an open telemetry standard led by the CNCF that unifies the data formats and transmission protocols for traces, metrics, and logs.
65%UnverifiedTraces are a core concept in distributed system observability, originating from standards like OpenTelemetry
90%UnverifiedOpik支持嵌套Span结构进行多层级调用可视化
90%UnverifiedFoundry的追踪系统基于OpenTelemetry标准构建
90%UnverifiedOpenTelemetry等标准化追踪框架的设计假设是确定性的调用链路，不适合AI Agent的非确定性执行路径
80%UnverifiedThe OpenTelemetry project standardizes the concept of distributed tracing, defining a unified Trace data model and propagation protocol.
70%