Do you need access to my source code?

No. NSPEC tests the running UI. You give it a URL and optional login. It never reads your repo unless you opt in to git-diff based risk prioritization.

Which viewports are covered?

Six viewports at launch: desktop 1440, laptop 1280, tablet portrait and landscape, mobile portrait and landscape.

Yes, on Enterprise. Docker and Helm, with BYO LLM (OpenAI, Anthropic, or a local model). Artifacts never leave your network.

Now accepting early access · Q3 2026

Point it at a URL.Get verified bugs back.

Q: How does NSPEC verify bugs?

An independent bug-verifier agent re-runs the repro in a fresh browser context, up to three times. Only bugs that reproduce, with a manual-grade confidence score, make it into the report.

NSPEC is an autonomous frontend QA agent. It crawls your app across six viewports, runs eight specialist agents in parallel, verifies every finding independently, and files the tickets straight into Jira, Linear, or GitHub over MCP.

Join 487 engineers·Launching Q3 2026·No credit card

0agents

0viewports

0test code

nspec run

live

https:// running

8 agents : parallel

6 viewports

orchestratorscheduling 247 steps
ui-explorer/games 3/6 viewports
component-audit41 controls
responsivemobile_390x844
bug-verifierreproduce BUG-003

›14 steps · 0 bugs found02:41

report ready : bugs / BUG-003

P1 · verified

Impossible RTP values on game detail

screenshot

DOM

console

network

reproduced 1/3 : confidence high

third-party noise filtered

not a duplicate of prior runs

The state of frontend QA

Your tests go green. Then prod breaks.

Between flaky suites, 300-finding crawler dumps, and unhired QA, most teams ship and pray. NSPEC replaces the prayer.

suites flaky in 90 days

Record-and-replay rots.

Every release shifts selectors. The suite goes red, nobody trusts it, and real regressions get buried in the noise.

findings per run, 3 real

Crawlers cry wolf.

Generic QA tools flood you with third-party console errors, retries, and font 404s. Bugs that aren't yours. Fixes nobody ships.

viewports per route

Manual specs don't scale.

Between new routes, six viewports, and edge cases, hand-written coverage can't keep up. One manual pass misses what a real user hits.

What comes back

A ticket, not a report.

Verified findings land directly in Jira, Linear, or GitHub over MCP · title, repro, acceptance criteria, evidence bundle, owner. This preview is what the ticket body looks like.

bugs / BUG-003

› bug.json
› repro.md
› highlighted.png
› fullpage.png
› dom.html
› console.log
› network.json
› verifier.json

BUG-003P1verified · 1/3 reproconfidence: highqa-run/20260401/bugs/BUG-003

Impossible RTP values displayed on game detail page

Some games render RTP values greater than 100% or negative. The display contract expects a clamp to [0, 100]. Source evidence attached.

/games/golden-spin

Game detail · Golden Spin

volatility

medium

max win

5000x

min bet

$0.10

RTP

124.8%

highlighted.png

1440×900

dom.html

17 KB

console.log

12 lines

network.json

8 reqs

verifier · independent run

reproduced on attempt 1 of 3
manual confidence: high
third-party noise filtered
not a duplicate of prior runs

file via mcpJiraLinearGitHubNotion

How it works

Three inputs from you. Zero test code.

Point

Hand it a URL.

Paste a staging link, production URL, or preview deployment. Add optional login. That's the whole input.

Run

Eight agents, six viewports.

NSPEC crawls routes, tests controls, fills forms, and audits components in parallel. Each agent in an isolated browser. Desktop to mobile, covered.

Ship

Verified tickets, in your tracker.

A verifier agent reproduces every finding. Noise is filtered server-side. Survivors get filed straight into Jira, Linear, or GitHub over MCP · with a prototype HTML report attached for record-keeping.

What makes it different

Built to earn your trust on run one.

Every bug reproduced, or it doesn't ship.

A separate verifier agent re-runs each finding in a fresh browser up to three times. Only survivors get filed, each scored with a manual-grade confidence.

verifier.json

{

"bug_id": "BUG-003",

"attempts": 3,

"reproduced": true,

"confidence": "high",

"verdict": "shipped"

}

Eight agents, six viewports, one flat run time.

Agents work in parallel with isolated browser contexts. A 247-step coverage pass lands in under four minutes · without you writing or maintaining a single selector.

8 agents · parallel02:41 elapsed

orchestrator
100%
ui-explorer
92%
component-auditor
78%
responsive-tester
64%
performance-profiler
45%
bug-verifier
31%
test-case-designer
24%
accessibility-reviewer
18%

Tickets, not a report. Filed via MCP.

Verified bugs go straight into Jira, Linear, or GitHub Issues over native MCP connectors · title, repro, acceptance criteria, evidence, owner. The HTML report is a prototype preview, not the delivery channel.

filed via mcp3 filed

JiraWEB-4218Impossible RTP on game detailP1
LinearFE-902Table overflow · tabletP2
GitHub#314Duplicate product in filterP3

Third-party noise, filtered at the source.

Server-side gates drop empty findings, duplicate screenshots, GA/Sentry/ad-tech errors, and inter-agent contradictions. You see what matters. Nothing else.

findings : before gates

×GA collect 403
×Sentry envelope rate-limited
×font woff2 404 (fonts.example.cdn)
×hydration warning ×184
×adtrack cors blocked
×duplicate screenshot ×26
✓impossible RTP on game detailP1
✓tablet overflow · /games tableP2

Planned pricing

Priced per project. Not per seat.

Pricing finalizes at public launch. Waitlist members get early access before anyone else.

Solo

Freefor hobby projects

Enough to kick the tires on a side project.

3 runs per month
Standard tier · 4 viewports
Self-contained HTML report
Community support

The answers engineers ask first.

Every finding is handed off to an independent bug-verifier agent that re-runs the repro in a fresh browser context, up to three times. Only bugs that reproduce, with a manual-grade confidence score, make it into the report.

Be first in line

Your next QA pass
runs itself.

Join 487 engineers on the waitlist. Invites roll out in cohorts before public launch.

Launching Q3 2026·No credit card·No spam