Live instrument

Every feature, live —
and the spec that proves it.

Each card below shows Scout verifying the feature in a short recorded demo, a working surface you can exercise yourself, and the exact .scout.md scenario behind it. The spec on the page is the spec that runs — one file, no drift.

// the proof

Scout runs these very cases in CI

On every push and pull request, GitHub Actions builds this Lab fresh and replays the exact .scout.md scenarios below against it — deterministically, with no AI. Open the latest run to see Scout verify each one.

View the latest Scout run on GitHub

// features

Multi-tab follows

It follows the click into a new tab

A click opens a booking tool, an OAuth window, a report — and the agent follows. Scout records the tab switch as a deterministic step, then runs every assertion on the active tab.

Recorded with scout --demo-video. Press play to watch the scenario run.

The recording captures the first tab only — the new tab opens off-camera, so this clip ends as the click fires. Scout still follows the switch and runs every assertion on the new tab; see the spec below.

Try it

Opens /lab/report/ in a new tab — the agent follows the switch and asserts on the new tab.

multi-tab.scout.md the scenario

---
feature: Multi-tab follows
tags: [lab]
---

## A new tab opens with the report ready

On the Lab page, click "Open the report in a new tab". A new tab opens and shows the heading "Report ready". No errors appear in the browser console.

Read the docs

Environment interpolation

Secrets stay in the environment

Write $ENV:VAR where a value goes; Scout resolves it from the environment at run time — in form fills and navigation URLs alike. The real value never enters the committed script and never reaches the model.

Recorded with scout --demo-video. Press play to watch the scenario run.

Try it

No coupon applied.

Scout types $ENV:LAB_COUPON here and asserts the coupon was accepted — the literal value never enters the spec, the recorded script, or the model.

env-form.scout.md the scenario

---
feature: Environment interpolation
tags: [lab]
---

## A coupon from the environment applies

On the Lab page, type $ENV:LAB_COUPON into the "Coupon code" field and apply it. The page confirms the coupon was applied — assert "Coupon applied" is visible. The literal value stays in the environment, so it never appears in this spec or the recorded script.

Read the docs

Network assertions

It checks the calls the page makes

Describe the expectation in prose — a POST returns 2xx with an orderId. Scout records a tolerant assertion matched on method, URL pattern, and status class, deliberately ignoring volatile ids so replay stays green.

Recorded with scout --demo-video. Press play to watch the scenario run.

Try it

No order placed.

Fires GET /lab/api/order.json — Scout asserts 2xx with an orderId.

network.scout.md the scenario

---
feature: Network assertions
tags: [lab]
---

## Placing an order fires the order API cleanly

On the Lab page, click "Place order". A GET to /lab/api/order.json returns 2xx and the response includes an orderId. No errors appear in the browser console.

Read the docs

Console assertions

It reads the browser console

Assert that a specific log appeared — or that no errors did. "No errors" covers console.error and uncaught exceptions; a positive assertion matches a stable prefix so unrelated console noise never makes it flaky.

Recorded with scout --demo-video. Press play to watch the scenario run.

Try it

Open your browser console, then click.

Scout asserts a console log containing DIAG:[scout-lab] appeared.

console.scout.md the scenario

---
feature: Console assertions
tags: [lab]
---

## Diagnostics log the calibration marker

On the Lab page, click "Run diagnostics". A console log containing "DIAG:[scout-lab]" appears. No other errors appear in the browser console.

Read the docs

Auth profiles

It runs as a signed-in user

Capture a logged-in browser session once per environment, then pick a profile per scenario. Authenticated flows replay from the saved session — no login step re-runs, no credentials in the script.

Recorded with scout --demo-video. Press play to watch the scenario run.

Try it

Scout captures this session once with scout login, then replays the qa profile.

auth.scout.md the scenario

---
feature: Auth profiles
profile: qa
tags: [lab]
---

## A signed-in user reaches the calibration panel

On the Lab page, sign in. Once authenticated, the calibration panel shows "Signed in as qa".

Read the docs

// quickstart

Point your own Scout at it

Run Scout against a known-good target

This page is a stable target. Point Scout at it with one npx command — no install — and watch it verify a scenario against a site you did not write. The fastest way to see a first reading.

Read your first scenario

$ npx @pcamarajr/scout create "A new tab opens with the report" -f demo -c "Open the report in a new tab; the new tab shows Report ready"$ npx @pcamarajr/scout go --base-url https://scout.pcamarajr.dev/lab

Every feature, live —and the spec that proves it.

Scout runs these very cases in CI

It follows the click into a new tab

Secrets stay in the environment

It checks the calls the page makes

It reads the browser console

It runs as a signed-in user

Run Scout against a known-good target

Every feature, live —
and the spec that proves it.