Trace LLM calls and AI agents with OpenTelemetry. Inspect token usage, cost, latency, and errors with TypeScript, Python, or an existing OTLP integration.