npm v1.8.3JS/TS livelocal scanApache-2.0

The evidence agent for AI-generated code

Verglos scans the vulnerabilities AI agents keep leaving in your code.

The free local scanner is live today. The v2 track adds local hunt verification and signed evidence so you can hand security proof to a client.

$ npx verglos

scanned 1,247 files in 42s

report: verglos-report.html + verglos-report.json

verglos scan

critical / D4-004

credential-shaped value in source

high / AI-005

unresolved package name in lockfile

fix

move to env, rotate, pin dependency

0/ 100

Critical risk

AI-authored estimate41%
Evidence artifactlocal

1.8.3

latest npm

local scanner live

2.0.0-alpha.1

source alpha

hunt + attest shells

3

moats

provenance · local-first · evidence track

0

source upload

scan runs on your machine

300

repo campaign

v2 methodology in progress

Why the market changed

Agentic AppSec is now table stakes.

ZeroPath, Corgea, Snyk Agent Fix, Cycode Maestro, XBOW, Semgrep AI Assistant, and Pixee all point at the same market shift: scanners are becoming triage-and-fix agents.

The shared pattern is cloud LLM triage plus PR fixes. That is useful, but it still asks the buyer to trust inferred severity and generated remediation.

Verglos takes a different lane: local sandbox proof, AI-authorship provenance, and signed evidence artifacts. It is built to prove what fired, show where agent-authored code concentrated risk, and package the result for handoff.

The Verglos flow

One release decision. Every proof point connected.

Verglos keeps source and scanning local, then connects its findings with evidence from Trivy, Docker/OCI, and the tools your team already runs. The result is a decision you can explain, not another isolated alert.

Hunt and signed attestations are shown as the v2 beta-track destination.

01

Input

AI-assisted code

Developer, agent, repository, CI

02

Inspect

Verglos scan

JS/TS findings, provenance, score

03

Enrich

Evidence sources

Trivy · Docker / OCI · existing tools

04

Decide

Correlate + policy

Normalize, deduplicate, pass / review / block

05

Prove

Release record

Portable evidence, monitoring, verification

Customer-controlled execution: code, images, scans, and ordinary reports stay local.

Optional hosted memory: history, monitoring, alerts, policy, and team context.

Verglos owns identity, correlation, policy, and the release record.

SAST + SCA · 10 domains

Built from the bugs AI agents make cheap to create.

One CLI runs SAST (code-side detectors — secrets, injection, misconfig, AI patterns) and SCA (dependency CVEs, vendored libraries, slopsquat / typosquat) across ten explicit domains: injection, auth, access control, cryptography, misconfiguration, supply chain, API surface, data exposure, runtime, and monitoring.

Domain 01 of 10

Injection & input flaws

Critical

Every piece of user input is a weapon if it reaches a dangerous sink unsanitised. We trace source-to-sink across your codebase — SQL, NoSQL, command exec, XSS, SSTI, path traversal — and flag the concatenations, spreads, and eval calls that turn a request into a takeover.

SQL / NoSQL
Command exec
XSS (stored / reflected / DOM)
SSTI

sample finding

element.innerHTML = req.query.name

app/user/[id]/page.tsx:42

Three moats

Verglos is not trying to win the cloud triage race.

AI-authorship provenance

Every finding is annotated human / AI / mixed with commit-shape and agent-artifact evidence.

Cloud scanners can triage code, but they usually do not preserve authorship evidence as a first-class risk dimension.

Local-first

Scan runs 100% on the developer's machine today. Hunt is designed for local verification so source does not need to leave the machine.

Cloud LLM products inherit cloud cost and source-upload friction; Verglos keeps the security loop local.

Signed evidence artifact

The v2 attest track is built around a portable JSON+HTML bundle signed with Ed25519 and anchored to a public verify URL.

A PR fix is not a client-portable evidence artifact. Studio is built around the handoff surface.

Evidence path

Enterprise thinking without pretending to be enterprise theater.

Verglos is designed to move from local findings to durable proof: CI enforcement, monitoring, signed attestations, verification URLs, SBOMs, and client-ready reports.

01

Scan

Deterministic pass, findings, confidence, and provenance in seconds. Local and free.

02

Hunt

v2 beta track

Synthesize a proof-of-exploit per finding, fire it in a local sandbox, and mark verified true, false, or not_attemptable. Pro track.

03

Attest

v2 beta track

Ed25519-sign the verified subset into a portable bundle with a public verify URL for client handoff. Studio track.

WHAT VERGLOS IS NOT

Not another autonomous pentester. Verglos verifies findings on code you own.

Not a cloud SAST. Verglos does not upload source to run an LLM triage pass.

Not a Vanta replacement. Verglos produces evidence artifacts, not the compliance program.

Not AI-powered as a headline claim. The scanner is deterministic; the hunt loop uses local synthesis and sandbox execution.

Why it matters

The invisible cost you're paying today.

Not the bugs Verglos finds — the ones you don't. Every AI-assisted stack carries a handful of buried failure modes that won't announce themselves. Here's what changes when you can see them.

$29/mo · less than one hour of dev time

The developer

You stop needing to remember.

The scan you meant to run last Tuesday is the one that would have caught the CVE that shipped today. Verglos watches your dependency tree hourly and emails the moment a critical is published. No calendar reminder, no manual re-scan, no discipline debt.

The agency

Client handover without the awkward Zoom.

Every AI-built app you ship carries the same buried defaults — wildcard CORS with credentials, jwt.verify accepting alg=none, MD5 password hashes. Your client's security review will find them. Better they surface at build time on your machine than in a pen-test report three months in.

The founder

The 3am email you don't want but need.

Log4j, event-stream, whatever ships next — critical CVEs land at 2am UTC on a Sunday and your customer notices before you do. Verglos flips that: an email arrives within the hour of publication, deduped so you get it once, delivered through email, Slack, or webhook. Cheaper than one bad Monday.

The agent user

AI can't ship the same bug twice.

Your coding agent will write Math.random() for a session token because a tutorial did. It'll suggest a package name because it hallucinated one. Verglos plugs into Cursor, Claude Code, Windsurf, and Cline as an MCP tool — the agent asks Verglos before writing risky code or installing a suspect package. Correction happens inside the flow, not after the merge.

Language coverage

JavaScript and TypeScript first.

JS/TS is live today because that is where Verglos has real detector coverage, package analysis, and provenance signal. Other stacks stay on the request list until they are actually supported.

Live now

JavaScript and TypeScript first.

JavaScriptJS
TypeScriptTS

Covers Node.js, Next.js, React, Express, Fastify, and npm workspaces.

PythonGoRubyJavaPHPRustC# / .NETSwiftKotlin
Request your stack →

Plans

Free is the complete local scanner.

Run it as many times as you want locally. Paid tiers are for teams that need enforcement, monitoring, evidence, handoff, and readiness workflows around the scan.

Free plan

Unlimited local scans. Full findings. Full paths. Local HTML and JSON reports. No account required.

Live

Free

$0

For: Developers, founders, agencies

Use: OSS scanner, provenance, and MCP scan.

Unlimited local scans. No account, no source upload, no paywall for findings, paths, or line numbers.

  • /Full SAST + SCA scan across 10 domains
  • /Security score /100
  • /All file paths + line numbers
  • /SAST: secrets, entropy, injection, misconfig, AI patterns
  • /SCA: dependency CVEs via OSV + vendored-library CVE cross-ref
  • /AI-authored code detection
  • /Slopsquat / hallucinated-package check
  • /Local HTML + JSON report
  • /Pre-commit hook
  • /MCP verglos_scan tool
  • /CI blocks on criticals
  • /Pre-commit hook
  • /Unlimited local use
Run from npm

V2 alpha track

Pro

$29/mo

For: Serious solo devs and small teams

Use: Hunt track, CI discipline, and agent guardrails.

For developers whose coding agent should verify high-signal findings before every commit. Functional hunt verification ships on the v2 beta track.

  • /Everything in Free
  • /verglos hunt shell for Critical + High
  • /MCP verglos_hunt_finding shell
  • /MCP verglos_hunt_report shell
  • /MCP verglos_hunt_before_write shell
  • /MCP verglos_hunt_explain_verdict shell
  • /verglos fix — provenance-annotated PR diffs
  • /verglos ci --hunt path for verified criticals
  • /CI score thresholds
  • /Continuous CVE monitoring
  • /Offline-safe license (Ed25519 signed)
Start Pro

Validation

Studio

$199/mo

For: Agency owners and delivery leads

Use: Attestation, client handoff, and proof resale.

Agencies do not need another noisy scanner. They need signed evidence they can hand to clients.

  • /Everything in Pro
  • /verglos hunt on Medium
  • /Signed attestations — shipping with v2.0.0-beta
  • /Public verify URL — shipping with v2.0.0-beta
  • /White-label report — shipping with v2.0.0-beta
  • /Firecracker sandbox adapter — shipping with v2.0.0-beta
  • /Agency dashboard — shipping with v2.0.0-beta
  • /MCP verglos_attest
Discuss Studio

Contact sales

Enterprise

Custom

For: Security-led organizations

Use: Enterprise controls around the evidence chain.

For teams that need organization-level administration, private verification infrastructure, and custom detector coverage.

  • /Everything in Studio
  • /SSO / SCIM
  • /Self-hosted verify chain
  • /Audit log
  • /Custom detector packs
Contact sales
CapabilityFreeProStudioEnterprise
verglos scan (all detectors)YesYesYesYes
Provenance layer + slopsquat + typosquatYesYesYesYes
Local HTML + JSON reportYesYesYesYes
MCP verglos_scan toolYesYesYesYes
verglos ci blocks on any criticalYesYesYesYes
Pre-commit hookYesYesYesYes
verglos secrets / deps / scoreYesYesYesYes
verglos hunt on Critical + HighNoAlpha shellBetaBeta
verglos hunt on MediumNoNoBetaBeta
MCP hunt toolsNoAlpha shellBetaBeta
MCP verglos_hunt_before_writeNoAlpha shellBetaBeta
verglos fix (provenance-annotated PR diffs)NoYesYesYes
verglos ci --huntNoAlpha pathBetaBeta
CI score thresholdNoYesYesYes
Continuous CVE monitoringNoYesYesYes
verglos attest (Ed25519 signed bundle)NoNoBetaBeta
Public verify URLNoNoBetaBeta
White-label reportNoNoBetaBeta
Firecracker sandbox adapterNoNoBetaBeta
Agency dashboardNoNoBetaBeta
SSO / SCIM · self-hosted verify · audit log · custom detectorsNoNoNoContact sales

Top Notchh operating stamp

Built by a product company, not a scanner content farm.

Top Notchh Solutions builds practical software products from real operational pain. Verglos is the security-evidence product in that portfolio: specific, local-first, honest about maturity, and built to become infrastructure for AI-assisted delivery.

Founder-led product judgment
Reusable scanner infrastructure
Evidence before decoration

From the Journal

Research and product notes.

Original research from the Verglos scan campaign, positioning notes, and product decisions — published on the Top Notchh Journal.

Read the full Journal →

FAQ

Boundaries make the product stronger.

Is Verglos just another scanner?+

No. The scanner is the entry point. Verglos is built around evidence: AI-code risk signal, report artifacts, MCP guardrails, CI enforcement, and client-ready proof.

Does Verglos upload source code?+

The local scanner writes HTML and JSON reports beside your code. Networked flows are explicit: login, activation, monitoring registration, and future hosted evidence.

Which languages are supported today?+

JavaScript and TypeScript are live. Other language icons are request-list indicators, not shipped coverage.

Who is behind Verglos?+

Verglos is built by Top Notchh Solutions, an AI-native product company building practical software products from real operational problems.