Skip to main content

​
1. Install and initialize

Inside the Next.js app project:
init creates stacknext.config.json and an instrumentation.ts or instrumentation.js file when they do not exist. It never overwrites an existing instrumentation file. See installation for the manual patch, other package managers and monorepo details.

​
2. Optional: configure your AI coding agent

This adds or updates only a marker-delimited StackNext block in AGENTS.md. Existing instructions are preserved. The block points agents at the local documentation shipped with the installed packages. See AI agent setup.

​
3. Start the collector and app

dev starts the collector, runs the project’s dev script, and writes .stacknext/run.json with a short-lived bearer token. Open the printed dashboard URL, usually http://127.0.0.1:4400, and make a request to the app.
The collector binds to loopback, so the dashboard is reachable from your machine and nowhere else. The in-memory sample stays empty until the app handles a request.

​
4. Verify

The Overview page should show requests, and coverage should report what the sample observed. not-observed and unknown are valid evidence states, not errors. StackNext Overview after a first request: request and error counts, p95 latency, recent request latency and diagnostic findings StackNext Overview after a first request: request and error counts, p95 latency, recent request latency and diagnostic findings

​
5. Inspect the sample

  • Requests and Traces show route, status, duration and span timelines.
  • Database shows query events; Drizzle durations are null.
  • Errors shows redacted errors and any correlated trace.
  • Diagnostics shows deterministic findings. Read the evidence and limitations before changing code.
  • Snapshots lets you freeze the current sample.
  • Runs, Baselines and Comparisons hold Playwright-linked regression evidence.
CLI equivalents:

​
Next steps