Sentry For AI and LLM Observability

Agents, LLMs, vector stores, custom logic—visibility can’t stop at the model call.

Video thumbnail

works with

Vercel AI logo
Vercel AI
OpenAI logo
OpenAI
Node.js logo
Node.js
Next.js logo
Next.js
SvelteKit logo
SvelteKit
Nuxt logo
Nuxt
Astro logo
Astro
Remix logo
Remix
SolidStart logo
SolidStart
Express logo
Express
Fastify logo
Fastify
Nest.js logo
Nest.js
Hapi logo
Hapi
Koa logo
Koa
Connect logo
Connect
Hono logo
Hono
Bun logo
Bun
AWS Lambda logo
AWS Lambda
Azure Functions logo
Azure Functions
Google Cloud Functions logo
Google Cloud Functions
Electron logo
Electron

Tolerated by 4 million developers

Github logo
Disney logo
Atlassian logo
Vercel logo
Airtable logo
Duolingo logo
Cloudflare logo
Slack logo
Metronome logo
Autodesk logo
Microsoft logo
Instacart logo
lyft
Miro logo
Monday logo
Cursor logo
Anthropic logo

Know When Your AI App Breaks (Before Your Users Do)

Silent LLM Failures: Catch Outages and API Errors Instantly

LLM endpoints can silently fail with downtimes, transient API errors, or rate limits. Imagine your AI-powered search, chat, or summarization just stopping without explanation.

Sentry monitors for these failures and alerts you instantly, so you can fix them before your users are impacted.

Learn about LLM Monitoring
Stay Available: Real-Time Alerts for LLM Issues

Stay Available: Real-Time Alerts for LLM Issues

When AI features break (wrong prompts, missing inputs, bad responses, etc) Sentry sends you an alert right away, and gets you to the line of code, the suspect commit, and the developer who owns it.

With support for providers like OpenAI, Vercel AI, and most custom setups that use OpenTelemetry, you can fix issues fast, before they hit more users.

Meet Seer: AI debugger that troubleshoots issues with 94.5% accuracy

You're already juggling complex prompts, unpredictable model output, and edge cases you didn’t even know existed. Now something’s breaking, and you’re stuck guessing.

Seer, Sentry’s AI-powered debugging agent, analyzes your error data to surface root causes fast (with 94.5% accuracy), so you can spend less time digging and more time fixing. It’s our AI... fixing your AI.

More About Seer
Full-Stack Coverage: Monitor More Than Just the LLM

Full-Stack Coverage: Monitor More Than Just the LLM

Bugs happen, whether it’s your UI, API, or the code connecting to your LLM. Sentry shows you exactly where things break, with full context and the replay of a user engaging with your frontend.

So when bad data hits your LLM, you know why, and where to fix it.

Learn About Replays

Tracing & Performance for AI Agents

Debug End-to-End: Full Agent Flows

When something breaks in your AI agent, like a model timing out or a tool failing silently, traditional logs miss the full picture.

Sentry shows the entire agent run: prompts, model calls, tool spans, and errors, linked to the user action that triggered it. You see what broke, why it happened, and how it affected the experience.

Learn About Tracing
Handle Unexpected Model Outputs: Make Sense of Bad AI Responses

Handle Unexpected Model Outputs: Make Sense of Bad AI Responses

Models sometimes return malformed JSON or outputs that don’t match what your app expects. These failures are hard to predict and easy to miss.

Sentry captures the raw model output, shows where it was generated, and traces how it broke downstream. So you can quickly debug the issue and build smarter fallbacks for unpredictable AI behavior.

Speed Up: Track and Optimize LLM Response Times

Speed Up: Track and Optimize LLM Response Times

LLMs can introduce unpredictable delays: one prompt returns in seconds, another takes much longer due to provider load or network issues.

Sentry shows you how your LLM calls are performing over time, with breakdowns by provider, endpoint, and prompt. It’s easy to spot slowdowns, debug performance issues, and keep your AI features fast and reliable for users.

Learn about LLM Monitoring

Understand and Optimize Your LLM Costs

Unexpected Spikes: Catch Costly Issues Early

A single large input or unexpected spike can drive up token usage fast.

Sentry gives you real-time visibility and alerts, so you can catch unusual patterns early and keep LLM costs under control.

Budget Protection: Real-Time Alerts for Usage Spikes

Sentry continuously tracks token consumption and LLM-related costs.

If usage patterns shift unexpectedly or costs begin to escalate, Sentry sends immediate alerts so you can investigate, pause, or throttle high-cost activity before it becomes a problem.

Optimize Spend: Analyze and Tune LLM Prompts

Optimize Spend: Analyze and Tune LLM Prompts

Sentry provides granular analytics on token usage and costs at the provider, endpoint, and even individual request level.

You can easily spot which queries, workflows, or features are consuming the most tokens, then dig into the details to optimize prompt design and trim waste.

Learn about LLM Monitoring

Getting started with Sentry is simple

We support every technology (except the ones we don't).
Get started with just a few lines of code.

Grab the Sentry JavaScript SDK:

Click to Copy
<script src="https://browser.sentry-cdn.com/<VERSION>/bundle.min.js"></script>

Configure your DSN:

Click to Copy
Sentry.init({ dsn: 'https://<key>@sentry.io/<project>' });

That's it. Check out our documentation to ensure you have the latest instructions.

"Sentry played a significant role in helping us develop [Claude] Sonnet"
Since adopting Sentry, Anthropic has seen:
10-15%

increase in developer productivity

600+

engineers rely on Sentry to ship code

20-30%

faster incident resolution

Learn more

Fix It

Get started with the only application monitoring platform that empowers developers to fix application problems without compromising on velocity.

© 2025 • Sentry is a registered Trademark of Functional Software, Inc.