CEDX Browser · IT & Security
CEDX Browser runs your specs across real browser fleets — desktop Chrome and Edge, WebKit mobile, headless bulk — and scores every script for flake, so a red run means the product broke, not the harness.
checkout.hold.spec passes 72% of the time with a flake score of 88 across three browsers. The lab calls it a quarantine candidate on its own Overview — before your team wastes a morning on it.
Runs on demo data — the Browser lab and its runs are the software's sample, not a customer.
What it is
The script table carries suite, pass rate, flake score, runs and heals: checkout.hold.spec at 72% pass and flake 88 is failing; wms.bridge.spec at 96% with flake 47 is the healthy end. Three specs sit quarantined — auto-skipped, not deleted.
The run table names the browser per run — Chrome 125, WebKit 16, Edge 126, Firefox 127, headless — with duration, flake and status: a loyalty spec timing out at the 5-minute hard cap, a checkout spec failing in 2.2 minutes with flake 88.
96 sessions in the sample: 39 live, 3,402 billable minutes, $256.46 across the sample, 14 disconnects. Cost per session is on the row — a 90-minute Android session at $7.20, a headless Linux run at $3.44.
Product tour
Not a mockup and not a concept deck. This is what opens at /app/browser.
01 — Overview
Pass rate 62.8%, p95 duration 5.0 minutes, $4.0K fleet spend month-to-date at demo scale. The root-caused alerts do the triage: a WebKit mobile p95 up 48 seconds against baseline, 14 concurrent sessions pushing the plan's slot limit on the nightly.
02 — Runs
Sort by flake and the story tells itself: the runs that fail with low flake are bugs; the ones that fail at flake 88 are the harness arguing with itself. Both are in the table, distinguished.
03 — Scripts
Average pass across the suite is 86.8% — but the cards lead with the 12 high-flake specs, the 3 quarantined and the 8 failing. Quarantine is a first-class state: the spec auto-skips until someone heals it.
04 — Sessions
Every session bills minutes and dollars against its fleet: desktop Chrome at $7.20 for 90 minutes, WebKit mobile at $6.64 for 83. Errors are a filterable state — 14 disconnects in the sample.
Who runs it
Roles, not references. We have no named customers yet, so nobody in these photographs is quoted, credited or claimed as one.
Owns the failing specs — a red run with low flake is a bug to file, and the table makes that call in one sort.
failing · 8 specs
Keeps the fleets and the slots: the nightly's 14-concurrent-session pressure and the WebKit p95 regression are theirs to fix.
slots · 100/218
Watches spend and quarantine: $4.0K month-to-date at demo scale, and no quarantined spec older than its heal count justifies.
fleet spend · $4.0K demoThe shape of it
Every figure below is legible in the captures above. Nothing here is a projection of your suite — it is the state of the demo data.
How it runs
A spec schedules against a fleet — desktop Chrome, WebKit mobile, headless bulk — and waits its slot; 100 of 218 are in use right now.
The session bills minutes and dollars, streams its result back, and the outcome lands with duration to the tenth of a second.
Flake accumulates per script across runs and browsers; at 55 the spec is flagged, at 88 it is a quarantine candidate.
A quarantined spec auto-skips — the pipeline stays green on evidence, not on deletion — until a heal moves it back to active.
One record
A browser session is a device, an identity and a flow when it runs — the lab's record is the estate's record, read from another side.
The desktop fleets are managed devices with posture scores — the lab's chrome-desk fleet and the endpoint fleet are the same machines.
CEDX NetworkA session's traffic is a flow with a policy — what the lab's browsers can reach is the same rule stack everything else obeys.
CEDX Inventory ITThe lab's devices carry asset tags and refresh risk — a five-year-old fleet laptop shows up in both ledgers.
CEDX MonitoringA p95 regression on the WebKit fleet is a golden signal there too — the alert and the lab card tell one story.
Limits
Finding this out on the third call is worse for you than reading it here, and worse for us.
Start
Open it right now
Your specs, your fleets
Browser with the rest of it
Questions
Yes. Every screenshot is a capture of the running build and you can open the same build at /app/browser. It runs on demo data — the lab and its runs are the sample.
A per-script score built from its history across browsers: how often it fails without the product changing. At 55 the script is flagged; at 88 — like checkout.hold.spec — the Overview names it a quarantine candidate outright.
Auto-skips the spec so the pipeline stops crying wolf, while keeping the script, its history and its heal count visible. Three specs are quarantined in the demo; none is deleted.
Cost per session is printed on the row — $7.20 for a 90-minute Android desktop-fleet session, $3.44 for a headless Linux run — and the fleet spend card is labelled demo scale. This page is not a price list.
No certification has been issued. What we can evidence about hosting, tenant isolation and session handling is written up on the security page.
Live build, demo data, no card. Then ask which of your specs would quarantine first.