Sentry Observability

Sentry helps you monitor errors, traces, latency, and token usage for AI applications. Sentry's Next.js and Node.js SDKs instrument AI SDK v7 through the native Node.js telemetry channel, so you do not need @ai-sdk/otel or registerTelemetry to send AI SDK traces to Sentry.

Sentry captures spans for generateText, streamText, model calls, tool calls, embeddings, reranking, token usage, and errors. Add a functionId with the AI SDK telemetry option to make traces easier to find and group in Sentry.

AI SDK v7 support requires @sentry/node, @sentry/nextjs, or another Sentry server SDK version 10.62.0 or later. The native telemetry channel is available in the Node.js runtime; use a Node.js route or server process for these traces.

Setup

If Sentry is already configured in your application, skip to Usage. Otherwise, install and initialize the Sentry SDK for your runtime before running AI SDK calls.

Next.js

The Sentry wizard can create the full Next.js setup for you. The snippets below show the server-side pieces relevant to AI SDK traces:

pnpm add @sentry/nextjs

Create or update your Sentry server config and enable tracing with tracesSampleRate or tracesSampler:

sentry.server.config.ts
import * as Sentry from '@sentry/nextjs';


Sentry.init({
  dsn: process.env.NEXT_PUBLIC_SENTRY_DSN,
  tracesSampleRate: 1.0,
});

If you are setting up Sentry manually in Next.js, make sure your instrumentation.ts loads the server config for the Node.js runtime:

instrumentation.ts
export async function register() {
  if (process.env.NEXT_RUNTIME === 'nodejs') {
    await import('./sentry.server.config');
  }
}

Node.js 

pnpm add @sentry/node

Initialize Sentry at the start of your server entry point and enable tracing with tracesSampleRate or tracesSampler:

instrumentation.ts
import * as Sentry from '@sentry/node';


Sentry.init({
  dsn: process.env.SENTRY_DSN,
  tracesSampleRate: 1.0,
});

The Vercel AI integration is enabled by default in Sentry server SDKs. If you have disabled default integrations, add it back explicitly:

Sentry.init({
  dsn: process.env.SENTRY_DSN,
  tracesSampleRate: 1.0,
  integrations: [Sentry.vercelAIIntegration()],
});

Usage

Use the AI SDK normally. No registerTelemetry(new OpenTelemetry()) call and no telemetry.isEnabled: true flag are required for Sentry.

app/api/generate/route.ts
import { anthropic } from '@ai-sdk/anthropic';
import { generateText } from 'ai';


export const runtime = 'nodejs';


export async function POST() {
  const result = await generateText({
    model: anthropic('claude-sonnet-4-5'),
    prompt: 'What is the weather in Tokyo?',
    telemetry: {
      functionId: 'anthropic-weather-demo',
    },
  });


  return Response.json({ text: result.text });
}

For a standalone Node.js script, wrap the AI SDK call in a Sentry span so the AI spans have a parent trace, then flush before the process exits:

index.ts
import './instrumentation';
import * as Sentry from '@sentry/node';
import { openai } from '@ai-sdk/openai';
import { generateText } from 'ai';


await Sentry.startSpan({ name: 'ai-sdk-demo' }, async () => {
  const result = await generateText({
    model: openai('gpt-5-mini'),
    prompt: 'Write a haiku about observability.',
    telemetry: {
      functionId: 'haiku-demo',
    },
  });


  console.log(result.text);
});


await Sentry.flush(2000);

After the request runs, open Sentry and look for gen_ai.invoke_agent, gen_ai.generate_content, and gen_ai.execute_tool spans in your traces.

Recording prompts and responses

By default, Sentry records metadata such as model, provider, latency, token usage, and errors. Prompt and response content are only sent if you opt in.

Enable content capture globally with Sentry's dataCollection.genAI options:

Sentry.init({
  dsn: process.env.SENTRY_DSN,
  tracesSampleRate: 1.0,
  dataCollection: {
    genAI: {
      inputs: true,
      outputs: true,
    },
  },
});

Or enable it for one AI SDK call with the telemetry option:

const result = await generateText({
  model: openai('gpt-5-mini'),
  prompt: 'Summarize this support ticket.',
  telemetry: {
    functionId: 'support-summary',
    recordInputs: true,
    recordOutputs: true,
  },
});

To opt out of Sentry AI telemetry for a sensitive call, set telemetry.isEnabled to false.

Resources