What Is OpenTelemetry GenAI? — AI Glossary

What it is: OpenTelemetry GenAI is the set of semantic conventions for instrumenting AI workloads under the OpenTelemetry standard. Version 1.40.0 (April 2026) covers LLM client spans, agent spans, events, and metrics. Provides vendor portability across Langfuse, Logfire, Honeycomb, Datadog, and others.
Who it is for: Developers and SREs building observability for AI applications who want vendor portability.
Best if: You want a short reference on what OpenTelemetry GenAI is and why it matters for AI observability.
Skip if: You don’t run AI workloads at scale yet. Want one practical AI workflow every morning? Subscribe to our free daily newsletter.

What is OpenTelemetry GenAI?

OpenTelemetry GenAI is the set of semantic conventions that extend the open OpenTelemetry observability standard to AI workloads. As of April 2026, GenAI Semantic Conventions 1.40.0 cover four areas: LLM client spans, agent spans, events (prompt and completion content), and metrics. LLM client spans exited “experimental” in early 2026 and are now stable; agent and framework spans are still labeled Development but stable in production use.

Why does OpenTelemetry GenAI matter?

Vendor portability. Instrument your agent once using OpenTelemetry GenAI conventions and you can swap from Langfuse to Honeycomb to Datadog to Logfire without re-instrumenting your code. Before the standard converged, every AI observability vendor had its own SDK and trace format — switching meant rewriting. The standard also means tooling (alerting, dashboards, eval pipelines) can be built once and shared.

How does OpenTelemetry GenAI work?

The highest-signal items to instrument: create_agent and invoke_agent lifecycle spans, execute_tool for tool calls, the required metric gen_ai.client.operation.duration, and the recommended gen_ai.client.token.usage. Claude Code ships native OTel support — set CLAUDE_CODE_ENABLE_TELEMETRY=1, point the OTLP exporter at any compatible platform, and you get model name, tool name, token counts, and duration recorded automatically (prompt content off by default for privacy).

Related terms

• AI agent
• Claude
• Tool use
• Large language model
• Harness engineering

Learn more on Beginners in AI

• AI Agent Observability Guide
• Harness Engineering for Beginners
• Claude Code Best Practices

Sources and further reading

• OpenTelemetry — GenAI semantic conventions (official)
• OpenTelemetry blog — GenAI observability state of the spec (April 2026)
• Anthropic — Claude Code OTel observability

Last reviewed: May 2026. AI terminology evolves quickly — verify specifics on the official source pages above.

You may also like

• AI Agent Observability Guide
• Harness Engineering for Beginners
• Claude Code Best Practices
• AI agent
• Claude
• Tool use
• AI Glossary