How DeepTracer Works — Zero to Monitored in 5 Minutes
Setup guide  ·  5 minutes

From just deployed
to fully monitored.

Three steps. One npm install. No DevOps, no config files, no dashboards to learn. Your AI agent starts watching the moment you push to production.

01
Install
Add the SDK
One package, one new file in your project. That's the entire setup. No config, no environment decisions, no wrong paths.
$ npm install @deeptracer/nextjs
instrumentation.ts
import { init } from "@deeptracer/nextjs"
export const { register , onRequestError } = init ()
02
Deploy
Push to production
Deploy your app like you always do. The agent boots automatically, connects to your project, and starts collecting events in real time.
Agent status Active
Events collected 2,847
Errors captured 3 grouped
Next health check in 4m 12s
03
Relax
Your agent takes over
Errors get caught, root causes get investigated, and you get a plain-English explanation with the fix — before your users notice anything.
Root cause found
STRIPE_WEBHOOK_SECRET undefined in production. Webhook validation failing since deploy #47.
Add env var to Vercel
Get started free → See pricing
01 Install 2 minutes

Two files.
That's the entire setup.

You don't need to configure anything, make any decisions, or read a long docs page. Run one command, create one file, and your agent is connected to your project.

No config file
init() reads your env vars automatically
Works on Node.js and Edge
Same file runs on both runtimes
Zero impact on startup time
Async, non-blocking initialization
Never crashes your app
SDK fails silently if misconfigured
Where do I get my API key?
Sign up free → create a project → copy your key from the dashboard. Takes 60 seconds. No credit card.
Get your key free
$ npm install @deeptracer/nextjs
added 1 package in 0.8s
1 package is looking for funding
run `npm fund` for details
$ ▌
// instrumentation.ts — create this in your project root
// Next.js picks it up automatically. No imports needed elsewhere.
import { init } from "@deeptracer/nextjs"
export const { register , onRequestError } = init ()
// That's it. Errors, logs, and traces start flowing automatically.
// next.config.ts — wrap your existing config
import { withDeepTracer } from "@deeptracer/nextjs/config"
export default withDeepTracer ({
// your existing Next.js config goes here
})
// Enables source maps + error grouping in production.
# .env.local — one environment variable
DEEPTRACER_KEY = dt_your_key_here
# For Vercel: Settings → Environment Variables → add the same key.
# For Railway / Render: add to your service environment.
# Browser SDK (optional — for client-side error capture):
NEXT_PUBLIC_DEEPTRACER_KEY = dt_your_key_here
Run in your project root
That's the install. Your agent is connected to your project. Time to deploy.
Next: Deploy →
02 Deploy Push your code
Vercel Deployment ● Building
12:47:02 Cloning github.com/yourname/myapp
12:47:04 Installing dependencies
12:47:09 Building Next.js app
12:47:23 ✓ Deployment complete
DeepTracer Agent Initializing
✓ SDK loaded · Node.js runtime
✓ API key verified
✓ Error capture active
✓ Agent online · monitoring started
0
events captured
0
errors tracked
$0.00
LLM spend today
First health check running…

Deploy like
you always do.

git push. Merge the PR. Click Deploy in Vercel. Whatever your workflow is — don't change it. The agent boots automatically when your app starts. No restart required. No environment to configure.

What gets captured automatically
Server errors & unhandled exceptions
Stack traces, request context, fingerprinted & grouped
Slow API routes
p50 / p95 response times per route, automatically
LLM API calls
Cost, tokens, latency — every OpenAI / Anthropic call
Log lines
info, warn, error from your existing console.log calls
Edge runtime & middleware
Same coverage on both Node.js and Edge runtimes
Your agent runs its first health check 5 minutes after boot. Every 5 minutes after that, 24/7.
Your agent is live. Events are flowing in. Now comes the part that's actually interesting.
See the AI in action →
03 Watch This is what you're paying for

You get
investigations,
not alerts.

Most monitoring tools send you a notification that says "500 error." You're still left staring at a stack trace at 2am, guessing what went wrong. DeepTracer's agent reads the stack trace, cross-references your logs, finds the root cause, and writes the explanation for you.

❌ Other tools
Alert: 500 error on /api/checkout
Error: Cannot read properties of undefined
Stack: at handler (/api/checkout.ts:47)
← Good luck figuring out why.
✓ DeepTracer
Root cause: STRIPE_WEBHOOK_SECRET undefined
Started: deploy #47 at 2:41 PM today
Impact: 23 failed checkouts so far
Fix: Add env var to Vercel → redeploy
Guardian mode: automatic investigations
On the Pro plan, your agent doesn't wait for you to ask. Every 5 minutes it checks your app's health. If it finds something wrong — error spike, LLM cost surge, slow p95 — it investigates automatically and sends you the report.
HIGH
Payment webhook failing silently
Triggered automatically · 3 minutes ago AI Investigation
Root cause
STRIPE_WEBHOOK_SECRET is undefined in your production environment. The webhook handler calls stripe.webhooks.constructEvent() which throws when the secret is missing. Next.js catches the error and returns 500 silently — no alert was sent by Vercel.
Evidence
14:41:03 ERROR stripe.webhooks.constructEvent — No webhook secret
14:41:03 ERROR POST /api/webhooks/stripe → 500 (23 times)
14:38:51 INFO Deploy #47 — new build deployed to production
Suggested fix
1 Go to Vercel → Settings → Environment Variables
2 Add STRIPE_WEBHOOK_SECRET from your Stripe Dashboard → Webhooks
3 Redeploy. Webhook failures will stop immediately.
23 failed checkouts
2m 19s time to investigate
0 user complaints (caught first)
That's the full flow. Install → Deploy → Your agent takes over. Free to start. Your first AI investigation is one deploy away.
Get started free → See pricing

Works with everything you already use

No new tools. No new dashboards. Plugs into your existing stack in minutes.

For Lovable / Bolt / v0 users
You built with AI. Now protect it with AI.
AI-generated apps have blind spots — missing error handling, unoptimized API calls. DeepTracer is the AI that watches the AI's work.
For Cursor / Claude Code users
Ship 10x faster. Sleep just as well.
Faster shipping means more things can break. DeepTracer is your production safety net — catches what slipped through, explains it, helps you fix it.
For freelancers
Know before your client does.
Monitor all your client projects from one dashboard. When something breaks, you get the root cause and fix before they even notice. Look like a hero.