@agnost/agent-mode SDK Skill

To instrument AI agents with OpenTelemetry in Node.js, you can use a single-file bootstrap setup to initialize telemetry and wrap your AI SDK calls with tracking functions. This coordinates identity propagation and OTLP export across supported frameworks.

Identity propagation for AI SDK telemetry traces works by using setAgnostContext() and agent.track() to attach user and session data to traces. This standardizes identity propagation across OpenAI, Vercel AI, and Mastra integrations using AsyncLocalStorage and OpenInference.

Yes, you can use OpenTelemetry instrumentation with Vercel AI SDK and Mastra. The Skill provides framework adapters like instrumentVercelAI and createMastraExporter to enable OTLP export, and gracefully handles scenarios when optional adapters are absent.

OpenTelemetry AI agent instrumentation works in serverless and edge deployments, including Express routes and serverless functions. It is applicable in Node.js environments using supported AI SDKs across these deployment targets.

You do not need a specific OpenTelemetry backend to export AI agent traces. The architecture is extensible and works with OpenTelemetry backends and a broad set of adapters without forcing a specific backend choice.

Limitations include graceful handling when optional framework adapters like OpenAI, Vercel AI, or Mastra are absent, meaning full telemetry requires those adapters. It is also restricted to Node.js environments using the supported AI SDKs.