Standards

Shared semantics for agent activity traces

Standards can make traces easier to correlate and exchange, but a standard span or context header does not automatically preserve principal identity, delegated authority, human approval, external settlement, or tamper evidence.

Direct answer

Standardization improves a common language; sufficiency remains use-case specific.

Tracevity evaluates which reconstruction concepts each standard documents and marks missing or unclear semantics explicitly. It does not turn adoption into a score or imply that an implementation conforms merely because it names the standard.

Published documentation profiles

Standards in the current Trace Map

Observed 2026-08-28: dedicated repository main branch, Development; related core Semantic Conventions 1.44.0.

OpenTelemetry GenAI Semantic Conventions

Development-status OpenTelemetry semantics for generative AI agents, model calls, retrieval, tools, plans, evaluations, and MCP activity.

1 capability families documented in the reviewed material
4 explicit not-documented, unknown, or conflicting findings

OpenTelemetry project / CNCF: Semantic Conventions for GenAI agent and framework spans Create GenAI agent and Invoke agent attribute tables
OpenTelemetry project / CNCF: Semantic Conventions for GenAI agent and framework spans Accepted agent and framework span attribute tables
OpenTelemetry project / CNCF: Semantic Conventions for GenAI agent and framework spans Agent span system instructions, input messages, retrieval, and data source attributes

Read the evidence profile

Observed 2026-08-28: current unversioned public specification; no stability or maturity label asserted in the reviewed pages.

OpenInference

An OpenTelemetry-based semantic convention for AI model, agent, retrieval, tool, graph, evaluation, and annotation traces.

2 capability families documented in the reviewed material
2 explicit not-documented, unknown, or conflicting findings

OpenInference project / Arize AI: OpenInference Semantic Conventions Session, Agent, Graph, and LLM semantic attributes
OpenInference project / Arize AI: OpenInference Semantic Conventions User attributes
OpenInference project / Arize AI: OpenInference Semantic Conventions Input messages, prompt template, and retrieval document attributes

Read the evidence profile

Observed 2026-08-28: Trace Context Level 1, W3C Recommendation dated 2021-11-23.

W3C Trace Context

A W3C Recommendation defining interoperable HTTP trace-context propagation through traceparent and tracestate headers.

1 capability families documented in the reviewed material
9 explicit not-documented, unknown, or conflicting findings

W3C Distributed Tracing Working Group: Trace Context Sections 3.2 and 3.3: traceparent and tracestate
W3C Distributed Tracing Working Group: Trace Context Abstract and specification scope
W3C Distributed Tracing Working Group: Trace Context Abstract and data model

Read the evidence profile

Observed 2026-08-28: MCP core specification 2026-07-28; related OpenTelemetry MCP spans are Development.

Model Context Protocol

A JSON-RPC protocol for connecting AI applications to tools, resources, prompts, and related capabilities, with W3C trace-context propagation.

0 capability families documented in the reviewed material
4 explicit not-documented, unknown, or conflicting findings

Model Context Protocol project: Overview - Model Context Protocol JSON-RPC messages, lifecycle metadata, and reserved _meta trace fields
Model Context Protocol project: Overview - Model Context Protocol Authorization framework and implementation guidance
Model Context Protocol project: Overview - Model Context Protocol Server features: prompts, resources, and tools

Read the evidence profile

Counts summarize documentation posture across the 12 Tracevity families. They are not a quality score, conformance result, or Tracevity-tested sufficiency finding.

Read alongside

Standards and vendor exports answer different questions.

Review the format map for documented representations, then inspect individual system profiles for capture, export, integrity, and privacy limitations.

Explore formats