Vygl is an AI-native ASPM (Application Security Posture Management) platform that combines seven scan engines — SAST (OpenGrep), SCA (OSV), Malicious Package Detection (GuardDog), Secrets Detection (Gitleaks), IaC Scanning (Checkov), Container Image Scanning, and Cloud Security Posture for AWS accounts — with AI-powered triage. It finds security risks across code, dependencies, secrets, infrastructure, container images, and live cloud accounts, then uses AI to identify false positives and suggest fixes.

What types of vulnerabilities does Vygl detect?

Vygl detects SQL injection, XSS, command injection, path traversal, and SSRF in source code (SAST); known CVEs in dependencies like Log4Shell and Spring4Shell (SCA); malicious packages — typosquats, install-script backdoors, obfuscated payloads, and compromised maintainer releases across npm, PyPI, Go, and RubyGems (Malicious Package Detection); hardcoded API keys, database credentials, JWT secrets, and private keys using 200+ detection rules (Secrets); misconfigurations like public S3 buckets and unrestricted security groups in Terraform, Kubernetes, Docker, and CloudFormation (IaC); vulnerable OS and application packages in built container images (Container Scanning); and live misconfigurations in AWS accounts — public buckets and snapshots, security groups open to the internet, root accounts without MFA, disabled CloudTrail, and Route 53 records still pointing at released Elastic IPs (Cloud Security Posture).

How does Vygl's AI triage work?

Every finding from every scan engine can be reviewed by an LLM. The AI classifies each finding as a true positive, false positive, or uncertain, assigns a confidence level (high, medium, low), explains its reasoning with evidence from the finding, and suggests a fix for confirmed issues. Teams can bring their own LLM provider — including a local model — or use Vygl's built-in provider.

Does Vygl access my source code?

It depends on how you scan. With the CLI, scans run in your own pipeline or on your own runner, and only findings are uploaded — the rule, severity, file and line, a short code snippet, and a fingerprint. Your repository and its git history stay where they are. With Managed Scans, Vygl clones the repository to a temporary directory for the duration of the scan and deletes it as soon as the scan finishes.

What CI/CD pipelines does Vygl support?

Vygl runs as a Docker container and works with GitHub Actions, GitLab CI/CD, and any Docker-compatible pipeline. It can also run locally for development. With Managed Scans, Vygl scans every push to connected GitHub, GitLab, and Bitbucket repositories without any CI/CD configuration. By default, Vygl reports findings without blocking your pipeline; you can opt into blocking with Policy as Code rules or --fail-on severity gates.

Does Vygl scan cloud accounts?

Yes. Connect an AWS account by deploying one CloudFormation stack that creates a read-only IAM role, and Vygl scans the live account on the schedule you choose — from every few hours to monthly — or on demand. You can also run the same scan from your own CI with vygl scan --scan-types cloud. Vygl checks 90+ curated controls across public exposure, network exposure, identity and access, encryption, logging, and secrets in cloud resources, and flags Route 53 records still pointing at Elastic IPs you have released — a subdomain-takeover risk that no code or IaC scanner can see. Vygl stores no cloud credentials: each scan runs on a short-lived session. AWS is supported today.

How is cloud scanning different from IaC scanning?

IaC scanning reads the Terraform, CloudFormation, and Kubernetes files in your repository and asks whether a template will create something insecure. Cloud scanning reads your running AWS account and asks whether what is deployed right now is insecure — catching console changes, drift, and problems created by a deletion, such as a DNS record left pointing at a released IP address. Most teams want both, and Vygl runs both through the same findings queue and AI triage.

Does Vygl scan container images?

Yes. Vygl scans container images from the CLI (vygl scan --image) or by connecting your container registry. It finds vulnerable OS packages and application dependencies across every image layer, correlates container CVEs with your source-code SCA findings, and generates per-image AI security briefs with base-image upgrade advice and priority fixes.

Does Vygl detect malicious packages and supply-chain attacks?

Yes. Vygl's malicious package detection engine, powered by GuardDog, scans your dependencies for supply-chain attacks across npm, PyPI, Go, and RubyGems — including typosquats and confusable names, malicious install or postinstall scripts, obfuscated payloads, suspicious binary execution, exfiltration to attacker infrastructure, and compromised maintainer releases. This is separate from SCA: SCA finds known CVEs in legitimate packages, while malicious package detection finds packages that are intentionally hostile.

How is malicious package detection different from SCA?

SCA (Software Composition Analysis) matches your dependencies against the OSV database of known CVEs in legitimate open-source packages — vulnerabilities like Log4Shell where the maintainers wrote insecure code. Malicious package detection looks for packages that should never have been published in the first place: typosquatted packages mimicking popular libraries, install scripts that exfiltrate environment variables, obfuscated code, and releases pushed by compromised maintainer accounts. Both engines run on the same dependency manifests and both flow through Vygl's AI triage.

Which package ecosystems does Vygl scan for malware?

Vygl's GuardDog-powered malicious package detection covers npm (JavaScript/TypeScript), PyPI (Python), Go modules, and RubyGems — the registries most frequently targeted by supply-chain attackers.

Which container registries does Vygl integrate with?

Harbor, Amazon ECR, GitHub Container Registry (GHCR), Google Artifact Registry, and Docker Hub. Harbor images are scanned on push via webhook; images in the other registries are rescanned nightly. You can also scan any image directly from the CLI without a registry integration.

Does Vygl support SSO?

Yes. Vygl supports Single Sign-On via Google, GitHub, GitLab, and Microsoft Entra ID (formerly Azure AD), letting teams sign in with their existing identity provider.

Vygl is free for open-source projects. Commercial access is currently invitation-only. You can request access at https://app.vygl.io/request-access or by emailing access@vygl.io.

What scan engines does Vygl use?

Vygl uses seven engines: OpenGrep for SAST (static code analysis), OSV for SCA (dependency vulnerability scanning), GuardDog for malicious package detection (npm, PyPI, Go, RubyGems), Gitleaks for secrets detection (200+ rules), Checkov for IaC scanning (Terraform, Kubernetes, Docker, CloudFormation), container image scanning for OS and application package CVEs across every layer, and cloud security posture scanning for live AWS accounts.

How does Vygl compare to other security scanning tools?

Vygl combines SAST, SCA, malicious package detection, secrets detection, IaC, container image scanning, and cloud security posture in a single platform (vs running Semgrep, Snyk, GuardDog, TruffleHog, Checkov, Trivy, and a separate CSPM). It has AI triage built in for automatic false positive identification, keeps your repositories in your environment when you scan with the CLI, offers one-click AI Security Brief reports, and does not block CI/CD unless you ask it to.

What integrations does Vygl support?

Source repositories via Managed Scans: GitHub, GitLab, and Bitbucket (cloud). Container registries: Harbor, Amazon ECR, GitHub Container Registry (GHCR), Google Artifact Registry, and Docker Hub. Cloud accounts: AWS. Code review: GitHub, GitLab, and Bitbucket pull request comments. Notifications: Slack, Microsoft Teams, email, and custom webhooks. AI IDEs via MCP (Model Context Protocol): Claude Code, Cursor, Windsurf, and any MCP-compatible client. Identity: Single Sign-On with Google, GitHub, GitLab, and Microsoft Entra ID. Vygl also exports CycloneDX SBOMs and SARIF.

What is Vygl's MCP integration?

Vygl provides an MCP (Model Context Protocol) server that lets developers query security findings, triage and verify issues with AI, check exposure to a CVE, search the dependency inventory, inspect container images, and check project health directly from AI-powered IDEs like Claude Code, Cursor, and Windsurf. Available MCP tools include get_needs_attention, search_findings, get_finding_detail, ai_verify_finding, triage_finding, get_security_posture, get_project_health, get_cve_exposure, search_dependencies, and list_container_images.

What is Vygl's AI Security Brief?

The AI Security Brief is a one-click feature that generates a comprehensive AI-powered security report for your project. It analyzes all open findings across scan types and provides an executive summary, priority actions (critical issues to fix first), and quick wins (low-effort, high-impact fixes).

AI-NATIVE ASPM PLATFORM

Your whole security posture, in one place.

SAST OpenGrep

SCA OSV

Malware GuardDog

Secrets Gitleaks

IaC Checkov

Images OSV

Cloud AWS

See and fix security risks across your code, dependencies, secrets, containers, and cloud accounts — with AI that verifies every finding, suggests fixes, and watches for new CVEs. One platform, built for growing teams.

AI Triage·Suggested Fixes·AI Security Briefs·CVE Watch·Alerts to PRs, Slack & Teams·MCP for AI IDEs·Policy as Code

Request Access See how it works

Scan Engines

Platform

24/7

CVE Watch

worker/jobs/rebuild.py:6

SAST · OPENGREP

RULE python.lang.security.audit.subprocess-shell-true

1import subprocess

2from pathlib import Path

4def trigger_rebuild():

5 cmd = "git pull origin main && make build"

6 subprocess.run(cmd, shell=True, check=True)

Scanner ⚠ HIGH · COMMAND INJECTION

AI reviewing · taint sources · subprocess args · call sites

✓ FALSE POSITIVE · CLEARED CONF 94%

cmd is a local string literal with no user input. No tainted source reaches subprocess.run, and the rule fires on shell=True regardless of taint — safe to clear.

SUGGESTED · ALLOWLIST RULE

rule: python.lang.security.audit.subprocess-shell-true path: worker/jobs/rebuild.py

Verified 0.7s Qwen Coder · Local → Posted to PR #1284

services/billing/pom.xml:39

SCA · OSV-SCANNER

CVE CVE-2021-44228 · Log4Shell · org.apache.logging.log4j:log4j-core

36<dependency>

37 <groupId>org.apache.logging.log4j</groupId>

38 <artifactId>log4j-core</artifactId>

39 <version>2.14.1</version>

40 <scope>runtime</scope>

41</dependency>

Scanner ⚠ CRITICAL · RCE · CVSS 10.0

AI reviewing · reachability · logger config · advisory range

! CONFIRMED · REAL CONF 99%

2.14.1 is inside the Log4Shell range (fixed in 2.15.0) and reachable through the JNDI lookups in billing/log4j2.xml — exploitable in production. Upgrade to 2.17.1, which also closes the follow-up CVEs.

SUGGESTED FIX

Verified 1.1s Qwen Coder · Local → Slack #sec-alerts

tests/fixtures/seed.py:5

SECRETS · GITLEAKS

RULE gitleaks:aws-access-token · entropy 4.32

1import pytest

3@pytest.fixture

4def mock_s3_client(monkeypatch):

5 monkeypatch.setenv("AWS_ACCESS_KEY_ID", "AKIAIOSFODNN7EXAMPLE")

6 yield mock_s3()

Scanner ⚠ CRITICAL · HARDCODED AWS KEY

AI reviewing · key validity · file context · usage scope

✓ FALSE POSITIVE · CLEARED CONF 98%

Value matches AWS's documented example credential AKIAIOSFODNN7EXAMPLE — non-functional, used in their public docs. File is under tests/fixtures/ and only loaded by pytest. No production exposure.

SUGGESTED · ALLOWLIST PATTERN

rule: gitleaks:aws-access-token path: tests/fixtures/**

Verified 0.5s Qwen Coder · Local → Auto-suppressed

route53 / acme.io

CLOUD · AWS

CHECK Dangling DNS · record points at a released Elastic IP

1zone acme.io # public

2record api-legacy.acme.io. A 52.29.14.201

3ttl 300

4eip not held by this account

5ip range AWS public pool · eu-central-1

6region global

Scanner ⚠ HIGH · SUBDOMAIN TAKEOVER

AI reviewing · record liveness · EIP ownership · zone delegation

! CONFIRMED · REAL CONF 96%

The zone is public and api-legacy still resolves to an address this account no longer holds. It is back in AWS's pool — whoever allocates it next serves content on your subdomain.

SUGGESTED FIX · DELETE RECORD

api-legacy.acme.io. 300 IN A 52.29.14.201

Verified 1.3s Qwen Coder · Local → Closes once removed

One platform. Every layer of your stack.

From the first line of code to the container you ship and the cloud account it runs in — and AI verifies every finding so your team can trust the results.

Source Code

SAST · OpenGrep

SQL injection XSS Command injection Path traversal SSRF Insecure deserialization

Dependencies

SCA · OSV

Log4Shell Spring4Shell CVE tracking CycloneDX SBOM 7+ ecosystems

Malicious Packages

Malware · GuardDog

Typosquats Install-script backdoors Obfuscated payloads Maintainer takeovers npm · PyPI · Go · RubyGems

Secrets

Secrets · Gitleaks

AWS keys DB credentials JWT secrets Private keys 200+ rules

Infrastructure

IaC · Checkov

Public S3 buckets Open security groups Over-privileged IAM Terraform · K8s · CloudFormation

Container Images

Image · OSV

OS package CVEs App package CVEs Layer attribution Source ↔ image correlation

Cloud Accounts

Cloud · AWS

Subdomain takeover Public buckets & snapshots Internet-open security groups Root without MFA CloudTrail off

AI VERIFICATION LAYER

Every finding from every layer flows through AI triage. Confidence scored. False positives cleared. Fixes suggested.

✓ Cleared

! Real + fix

AI that triages every finding.

AI reviews every finding with verdicts, confidence scores, and suggested fixes — so you spend less time on false positives.

app.vygl.io/findings/a3e7f…

High SAST Open

python.django.security.injection.sql-injection

User-controlled data is used in a raw SQL query, which could lead to SQL injection.

backend/views/users.py:42-44

35 from django.http import JsonResponse

36 from .models import User

38 @login_required

39 @require_GET

40 def search_users(request):

41 """Search users by name."""

42 query = request.GET.get("q")

43 users = User.objects.raw(

44 f"SELECT * FROM users WHERE name LIKE '%{query}%'"

45 )

46 results = [

47 {"id": u.id, "name": u.name}

48 for u in users

49 ]

50 return JsonResponse({"users": results})

AI Verification

AI-generated — verify manually before acting.

Analyze this finding with AI to determine if it's a true or false positive and get a suggested fix.

One click, one security brief.

A full project report on demand. AI reads every open finding and tells you what to fix first — and what you can clear in five minutes.

app.vygl.io/projects/acme-api

acme-api main

Findings Dependencies Security Brief Scans

Security Brief

Generate a comprehensive AI-powered security brief for this project. Analyzes all open findings across scan types.

Connect once. Vygl keeps scanning.

Connect a repository, a registry, or an AWS account. Vygl scans code on every push, images from your registry, and your cloud on your schedule — findings flow to your IDE, Slack, and dashboard.

Source Code

GitHub

GitLab

Bitbucket

Container Registries

Harbor

Amazon ECR

GHCR

Google Artifact

Docker Hub

Cloud Accounts

AWS Accounts

Vygl Scan Engine

SAST · SCA · Malware · Secrets · IaC · Images · Cloud

AI triage on every finding

Delivered to

AI IDE Agents Claude · Cursor · Windsurf

Slack summary + report links

Microsoft Teams channel alerts + report links

PR / MR Comments inline findings + AI triage

Dashboard real-time posture updates

Zero CI/CD Config

No YAML, no pipelines, no Docker setup. Connect your account and scanning begins on the next push.

Ephemeral Cloning

Source code is cloned to a temp directory and deleted immediately after scanning. Nothing persists.

Org-Wide Policies

Cloud-managed rules apply to every managed scan. Enforce security policy across all repos from one place.

A new CVE drops. You know where it lands.

Vygl monitors your SBOM against new CVE disclosures. When one hits a dependency you ship, your team knows before it's exploited.

24/7 Continuous Monitoring

4× Daily Scan Cycles

SBOM-Aware

Every dependency from your scans is automatically tracked against live CVE feeds. No manual package lists.

Instant Notifications

Slack, email, or webhook alerts fire as soon as a new CVE matches a package in your SBOM.

Upgrade Guidance

Every alert carries the affected range, CVSS, EPSS, CISA KEV status, and the version that fixes it.

CVE Alerts Live

Critical

CVE-2026-31245 2 min ago

Remote Code Execution in jsonwebtoken

jsonwebtoken <9.0.2 3 projects affected

Notified via

High

CVE-2026-28891 6 hours ago

Prototype Pollution in lodash.merge

lodash <4.17.22 7 projects affected

Medium

CVE-2026-27433 1 day ago

ReDoS in semver range parsing

semver <7.5.5 12 projects affected

Scan what's actually running.

Connect an AWS account with a read-only role. Vygl reads the live account for the misconfigurations your templates can't show — including the DNS records that quietly hand a subdomain to someone else.

IaC scanning asks

“Will this template create something insecure?”

Reads Terraform, CloudFormation and Kubernetes in your repo.

Cloud scanning asks

“Is what’s running right now insecure?”

Reads your AWS account: console changes, drift, and anything a deletion left behind.

01

Read-only, and provably so

One CloudFormation stack creates a role Vygl can assume only with an external ID generated for your organization. Nothing it grants can write, and object contents stay unreadable. Delete the stack and access ends.
02

No stored cloud credentials

Each scan runs on a one-hour session that is never written down. Rather keep it all in-house? Run vygl scan --scan-types cloud from your own CI.
03

Subdomain takeovers, caught

A Route 53 record still pointing at an Elastic IP you released lives in no repo and no template. Vygl flags it with the zone, the record, and the address.
04

Signal, not a compliance dump

90+ curated checks, deduplicated per resource, verified by AI, and closed automatically once fixed. A scan that can’t see your account fails loudly — it never reports “clean”.

Cloud Account Scanned 12 min ago

acme-production

AWS · 123456789012 · eu-central-1, us-east-1 · Every 6 hours

0 Critical 6 High 14 Medium 8 Low 91 checks · 7 families

High Subdomain takeover AI · Confirmed

api-legacy.acme.io. A 52.29.14.201

Elastic IP released — the address is no longer held by this account.

sg-0a41f7e2 · SSH open to 0.0.0.0/0 High

acme-exports · public via bucket policy High

management-trail · log validation off Medium

eu-central-1 · EBS default encryption off Medium

AWS today. Scans run on the schedule you pick, from every few hours to monthly, or on demand — and land in the same queue, triage, and alerts as the rest of Vygl.

Your security data, inside your IDE.

Query findings and verify issues from Claude Code, Cursor, or any MCP-compatible IDE — zero context switching.

Claude Code acme-api

~/projects/acme-api codex-1 full-auto

users.py

38 @login_required

39 @require_GET

40 def search_users(request):

41 """Search users."""

42 query = request.GET.get("q")

43 users = User.objects.raw(

44 f"...%{query}%'"

45 )

46 results = [{"id": u.id}

47 for u in users]

48 return JsonResponse(..)

AI Chat MCP

Works with any MCP-compatible IDE

Claude Code

Codex

OpenCode

Cursor

Windsurf

Any MCP Client

Fits into your workflow.

Vygl notifies your team in the tools they already use.

github.com/acme/acme-api/pull/247

feature/user-prefs main

Conversation 3 Files changed 4

vygl-bot bot commented 2 minutes ago

Vygl Scan Results

acme-api — feature/user-prefs | SAST, SCA, MALWARE, SECRETS, IAC

47 findings — 2 critical | 8 high | 12 medium | 3 low
5 new | 42 recurring

New Findings (5)

Severity	Type	Finding	File
CRITICAL	SCA	Log4Shell CVE-2021-44228 in log4j-core@2.14.1	`pom.xml`
CRITICAL	SAST	SQL injection via f-string in raw query	`backend/views/users.py`
HIGH	SECRETS	Hardcoded API key	`config/settings.py`
HIGH	SAST	Command Injection via os.system()	`backend/utils/deploy.py`
MEDIUM	SCA	CVE-2019-12384 in jackson-databind@2.9.8	`pom.xml`

Recurring Findings (42) — ...and 42 more

AI Triage Summary

True Positive Log4Shell — JNDI lookup enabled, log4j-core 2.14.1 is exploitable. Upgrade to 2.21+.

True Positive SQL Injection — user-controlled input flows directly into raw SQL query without parameterization

Likely FP Command Injection — input is validated through allowlist, os.system() receives only predefined commands

View full scan results

Posted by Vygl — security scanning with AI-powered triage

GitHub PRs

GitLab MRs

Bitbucket PRs

Non-blocking by default

Available

Slack

Available

Real-time scan results with severity breakdowns, new vs recurring findings, AI verdicts, and clickable links to every finding.

Scan completion summaries
Critical & high finding alerts
Direct links to findings
Test notifications

Microsoft Teams

Available

Get scan notifications and finding alerts delivered directly to your Teams channels.

Scan completion summaries
Critical & high finding alerts
Direct links to findings
Test notifications

Webhooks & Email

Available

Send scan and CVE events as JSON to any HTTP endpoint — a SIEM, an incident tool, your own automation — or straight to an inbox.

Per-event filtering
SSRF-safe delivery
Stable event IDs for dedupe
Test sends

Your whole security posture, in one place.

One platform. Every layer of your stack.

Your security posture, at a glance.

AI that triages every finding.

python.django.security.injection.sql-injection

One click, one security brief.

Security Brief

Connect once. Vygl keeps scanning.

Zero CI/CD Config

Ephemeral Cloning

Org-Wide Policies

A new CVE drops. You know where it lands.

SBOM-Aware

Instant Notifications

Upgrade Guidance

Scan the images you ship.

CLI Scanning

Registry Integration

Source ↔ Container Correlation

Per-Image AI Analysis

Scan what's actually running.

Read-only, and provably so

No stored cloud credentials

Subdomain takeovers, caught

Signal, not a compliance dump

Your models, your choice.

Prefer the command line? Run it anywhere.

Your security data, inside your IDE.

Fits into your workflow.

feat: add user preferences API endpoint

Vygl Scan Results

Slack

Microsoft Teams

Webhooks & Email

Security posture you can actually keep up with.