AI AppSec · runs inside your environment

Find business logic vulnerabilities
before they ship.

Runs inside your environment. 6 scanner engines plus AI cross-file reasoning — ranked findings, exploit paths, and patches. Source code never transmitted.

CWE·OWASP Top 10·CVSS v3.1·OWASP ASVS·NIST 800-53·PCI DSS·SOC 2
0.00%OWASP Benchmark detection
0%Fewer false positives
0 minTime to first finding
0 linesSource code transmitted
KritSecLLMbeta 0.3
Purpose-built security LLM — trained on CVEs, exploit chains, OWASP taxonomy, and breach patternsLearn more →
GA target Nov 2026

16 languages · 13 frameworks · 3 clouds · 6 IaC · 7 package managers

Intelligence pipeline

Six scanners in.
Enriched findings out.

Krita normalizes output from every scanner, deduplicates by SHA-256 fingerprint, then runs two AI layers — per-file deep scan and cross-file reasoning — before a single alert reaches your team.

01 · IngestINPUT
Git repositoriesURL / OAuth / API key
GitHub · GitLab · Bitbucket · GHES
CI/CD triggersPR metadata attached
GitHub Actions · GitLab CI pipelines
Scheduled scansconfigurable cadence
Cron-based recurring scans
Inbound webhooksHMAC-SHA256 verified
Any external system trigger
File uploadoffline / air-gapped
ZIP archive · tarball · direct path
API triggerAPI key scoped
POST /api/v1/scans with repo + branch
6 connectors active · forwarding to engine
Krita Enginecode-scan · v4.0
Pipeline nodes
SAST
SCA
Secret Detection
IaC
AI Deep Scan
Cross-File
Live run
clone acme-corp/payments-api
·discover 47 endpoints · 6 data owners
sast 38 findings
sca 12 vulnerabilities
secrets 3 exposed credentials
iac 7 misconfigurations
·dedup SHA-256 · 9 duplicates removed
deep_scan AI · 47 files · exploit scores
cross_file business logic · auth gaps
·aggregate 61 unique findings
persist tasks + SLAs + patches ready
03 · ExportOUTPUT
SARIFindustry standard
GitHub Advanced Security · GitLab Dashboard
Issue trackersauto-discovered fields
Jira · GitHub Issues · GitLab Issues
Alert routingCRITICAL ≤ 5 min
Slack · Teams · PagerDuty
Reportson-demand or scheduled
PDF · JSON · Generic outbound webhooks
PR commentsCI gate
Inline findings on GitHub · GitLab MRs
Outbound webhookany severity filter
Custom payload to any endpoint
6 destinations ready · receiving from engine
Every finding exits tagged → CWE · OWASP A-class · CVSS vector · NIST control · PCI DSS requirement
0
Attack surfaces
0+
Scanner engines
0
Compliance frameworks
0
Lines source transmitted
Verified findings

Verified against a real codebase.

Every finding below comes from a real scan of a public repository. No synthetic data, no hypotheticals.

276Findings
60CRITICAL
8Business logic rules
0Lines source transmitted
CRITICALSAST

Remote Code Execution via Insecure Deserialization

solutions/solutions.md:1-5

The node-serialize library's unserialize function allows deserialization of functions, enabling arbitrary code execution.

AI Fix

Use JSON.parse() for structured data. Avoid libraries that allow function deserialization.

CWE-502A08:2021
HIGHSECRETS

Authentication Bypass via Weak JWT Secret

config/auth.js:12

The JWT secret is set to a common dictionary word that can be cracked in milliseconds using Hashcat or John the Ripper.

AI Fix

Use a cryptographically strong random string of at least 32 characters stored in a secure environment variable.

CWE-522CWE-1391A07:2021
HIGHBUSINESS_LOGIC

All API endpoints missing authentication middleware

controllers/auth_controller.js:48

Cross-file reasoning identified that multiple API and GraphQL endpoints lack the authenticateToken middleware, exposing them to unauthenticated access.

AI Fix

Apply authenticateToken middleware to all routes except explicit public endpoints.

CWE-521A07:2021
KritSecLLM · beta 0.3

A reasoning layer built
for AppSec workflows.

General-purpose models apply generic reasoning to security problems and get generic results. KritSecLLM is purpose-built for AppSec — trained on CVE databases, exploit chains, OWASP taxonomy, and real-world breach patterns — so it understands authorization boundaries, data flows, and business logic the way your security team does.

KritSecLLM
AppSec · Business Logic
Live
Type
Security LLM
Version
beta 0.3
Reasoning
Cross-file · Multi-hop
Output
Findings + Patches
Capabilities · 20 active
Architecture mental modelData flow tracingAuth boundary analysisBusiness rule inferenceExploit path generationPatch synthesisIDOR detectionBOLA detectionPrivilege escalationMissing auth inferenceCross-service reasoningPII classificationSecret exfiltration pathsJWT / session analysisMulti-tenant isolationInput validation gapsAsync worker flawsOWASP ASVS mappingCWE auto-taggingCVSS scoring
Context: full codebase · cross-file reasoning active
KritSecLLM01·Understandacme-corp/payments-api
Repository structure
Django REST API · 47 endpoints · payment processing domain
PostgreSQL backend · Stripe payment gateway integration
JWT-based authentication across 14 protected routes
Celery async workers · Redis session cache
Building mental model of architecture and data ownership
Entry points mapped
·POST /api/payment/ → payments/views.py:34
·POST /api/webhook/payment → payments/webhook.py:12
·GET /api/admin/users/ → admin/views.py:88
Tracing data flows and ownership boundaries
Data flow model
·user_input → api/payment.py:18 → process.py:84 → cursor.execute()
·card_number, cvv classified as PII — tracked across 6 files
·Stripe gateway receives validated payment object from process.py
RESULT
Architecture model complete · 47 routes · 6 data owners · 3 async flows
Phase 01 · Understand|Layer 1: deep scan · Layer 2: cross-file
0/16
0.00%
OWASP Benchmark detection rate
0%
Reduction in false positives vs GPT-4
0%
Security training corpus — no general data
0
Compliance frameworks in training set
KritSecLLM is currently in active development — training runs ongoing.
RoadmapGeneral availability target — November 2026
Built for security teams

Everything in one platform

Scan, triage, assign, remediate, and report — without stitching together five tools.

AI Layer 1 + Layer 2

AI Deep Scan

Per-file vulnerability analysis with full AI enrichment — exploit scoring, CWE/OWASP mapping, and AI-generated patches. Two reasoning layers run in sequence.

Live scan output
AI Deep Scan · 47 files
auth/login.ts — CWE-89 · exploit: 9.1
payments/process.py — CWE-20 · exploit: 7.4
webhook.py — missing auth · CRITICAL
·api/admin.ts — IDOR pattern detected
Cross-file analysis
user_input → cursor.execute() traced
·card_number flows to 6 files unmasked
61 findings · 2 patches generated

Cross-File Reasoning

Traces execution paths across modules, middleware, and services. Finds IDOR, BOLA, and auth gaps that per-file analysis misses.

Alert Triage + SLA

Severity-based management with enforced SLAs — 4h CRITICAL, 24h HIGH, 72h MEDIUM. Auto-escalation and merge blocking on breach.

Compliance Mapping

Every finding auto-tagged to CWE, OWASP, CVSS, NIST 800-53, ASVS, and PCI DSS. Evidence artifacts generated for SOC 2, ISO 27001, HIPAA.

SARIF + PR Comments

Industry-standard SARIF export for GitHub Advanced Security, GitLab Security Dashboard, and Azure DevOps. Inline PR comments on every scan.

Integrations

Findings flow into your existing tools automatically — no manual export, no copy-paste.

Jira
Auto-create issues, map severity → priority, sync status
Slack
CRITICAL alerts within 5 min, channel routing by severity
PagerDuty
On-call escalation for CRITICAL findings
GitHub
PR comments, merge blocking, Actions trigger
GitLab
MR comments, CI pipeline integration, SARIF upload
Webhooks
HMAC-verified outbound events for any custom system

Deployment

Air-gapped by default in both modes. Source code never leaves your environment.

Enterprise On-PremYou own and operate the infrastructure. Air-gapped by default. Source code never leaves your network.
Krita Managed InfraWe operate the infrastructure for you. Still air-gapped by default — your data stays fully isolated.
92%
fewer false positives
45 min
scan to patch
6+
scan engines
6
standard families
CWE · OWASP · CVSS · NIST · ASVS · PCI
Not just another scanner

Sec. teams are drowning in noise.
Krita fixes the signal.

The problems below aren't edge cases — they're the daily reality for every security engineer. Here's what we do about each one.

The problem
How Krita solves it
Alert fatigue

78% of scanner alerts are false positives. Engineers spend days triaging noise instead of fixing real issues.

AI-verified findings

AI enrichment validates every finding against your actual execution paths before it reaches your queue. Only real issues surface.

Blind spots

Business logic flaws — IDOR, BOLA, privilege escalation — are invisible to pattern matchers. They ship completely undetected.

Cross-file reasoning

Krita builds a cross-file model of your authorization boundaries and ownership rules, then checks every route against them.

Tool sprawl

6 point tools, 6 different formats. Aggregating findings is a full-time job before any fixing can even start.

One unified pipeline

One pipeline ingests 8 attack surfaces and normalises every finding into a unified schema with CWE, OWASP, CVSS, and AI-generated fixes.

No context

A finding says 'SQL injection at line 47'. It doesn't say why it's exploitable in your architecture, or how to actually fix it.

Exploit path + patch

Every finding exits with an exploit path, a CVSS vector, an AI-written patch, and a task with an SLA — ready to assign in 45 minutes.

49%

of critical bug bounty findings are auth and business logic flaws — invisible to every scanner

78%

false positive reduction when AI verifies execution paths before surfacing alerts

0

lines of source code ever leave your environment — only normalized findings cross the boundary

From commit to fix

Connect a repo.
Leave with a fix.

Connect a repo. Get ranked findings with AI-generated patches, assigned tasks, and blocked merges — in one pipeline.

Stage 01

Link your repo. That's the whole setup.

No agents to install. No new infrastructure. Connect to GitHub or GitLab, and the pipeline runs in your existing CI environment. Your source code never leaves your network.

Outcome
Every commit scanned automatically. CRITICAL findings block the merge before they reach production.
acme-corp / payments-api · connected
acme-corp / payments-api
main · 1,247 files · TypeScript, Python · last commit 4m ago
Repository access confirmed
CI pipeline integration active
Branch protection rules loaded
Scan policy configuration ready
Ready to analyze

Every finding comes with an AI-generated patch, exploit path, and compliance mapping — ready to assign in 45 minutes.

8 attack surfaces

Every vulnerability class.
One pipeline.

From SQL injection to privilege escalation, from leaked secrets to cloud misconfigs — 8 attack surfaces covered with deterministic scanners and AI-enriched reasoning.

Static Application Security Testing
CRITICAL
SAST
16 languages
InjectionXSSPath TraversalAuth BypassInsecure DeserializationSSRFXXERace ConditionsMemory SafetyHardcoded Secrets
deep scan · acme-corp/payments-api
────────────────────────────────────
[CRIT] CWE-89 auth/login.ts:47
SQL injection via user_input → cursor.execute()
Exploit score: 9.1 | CVSS: 9.8
Fix: Parameterize query, add @validate_input
 
[HIGH] CWE-79 components/Comment.tsx:112
Reflected XSS in renderComment()
Exploit score: 7.6 | CVSS: 6.5
Fix: Sanitize output, use textContent
 
[OK] 5 findings · 2 patches · tasks created
────────────────────────────────────
KBusiness Logic Flaws
CRITICAL
Business Logic
AI-only
IDORBOLABFLAPrivilege EscalationMissing AuthTenant IsolationMass Assignment
Software Composition Analysis
HIGH
SCA
7 ecosystems
Known CVEsLicense ViolationsTransitive DependenciesOutdated PackagesMalicious Packages
Secrets & Credentials
HIGH
Secrets
git history
API KeysCloud CredentialsTokens & PATsPrivate KeysDB PasswordsWebhook Secrets
Infrastructure as Code
MEDIUM
IaC
CIS benchmarks
Public StorageOverpermissioned IAMOpen Ingress RulesUnencrypted ResourcesMissing Logging
Cloud Misconfiguration
HIGH
Cloud
AWS · Azure · GCP
IAM PostureNetwork ExposureEncryption GapsAudit LoggingMFA Enforcement
Network Vulnerability
MEDIUM
Network
65,535 ports
Exposed ServicesDeprecated ProtocolsCVE MatchingService FingerprintingCertificate Issues
Runtime Security
INFO
Runtime
event-driven
Container EscapesSuspicious ProcessesFile IntegrityLateral MovementPrivilege Changes
AI-native security

The flaws scanners can't find.
AI finds them by reasoning.

Half of critical bug bounty findings are authorization and business logic flaws — invisible to every rule database ever written. Krita uses AI to reason about your system, not just pattern-match against it.

KritSecLLM
Default · Purpose-built security intelligence
Active

Trained exclusively on security corpora — CVE databases, exploit chains, OWASP taxonomy, real-world breach patterns. Ships as the default reasoning layer. No prompt-engineering required, no hallucinated exploits.

AppSec · Business Logic
Cross-file reasoning
CWE + OWASP auto-tag
Exploit path synthesis
Or bring your own model

Model-agnostic by design. Swap in any provider for any reasoning task — no vendor lock-in, no data leaving your environment unless you choose it.

OAI
OpenAI
ANT
Anthropic
AWS
AWS Bedrock
AZR
Azure
OLL
Ollama
LGR
LangGraph
What AI reasoning detects
CRITICAL

Auth & authorization

The highest blast radius flaw class. A single missing check means an attacker doesn't need to exploit anything else — they just walk in. AI traces every permission gate, session mechanism, and JWT claim across all routes.

CRITICAL

Business logic flaws

IDOR, BOLA, privilege escalation. No rule database covers these — they require understanding what your application is supposed to do, then finding where it doesn't enforce that. Only AI with system-level context can reason about them.

HIGH

Cross-file data flows

Auth defined in service A, missing in service B. Input sanitized in the route handler, bypassed three layers deep. Cross-file reasoning maps the entire execution path — not just individual functions.

HIGH

AI-generated code risks

Vibe-coded vulnerabilities: hallucinated imports, insecure LLM defaults, copy-paste credential leaks. Purpose-built to catch the new class of flaws that ship when developers trust LLM output without review.

Source code never leaves your environment. The scan engine runs in your CI/CD or infrastructure. Only normalized finding JSON crosses the boundary.

Built for

Most scanners find what they already know how to find.

Krita reasons across your codebase — catching business logic flaws, cross-file vulnerabilities, and authorization gaps that pattern matchers miss. Inside your environment.

CS
CISO / AppSec lead

Reduce risk and audit overhead

Every finding auto-mapped to CWE, OWASP, and 43 compliance frameworks. Audit trail built in.

EL
Engineering lead

Block vulnerable merges before they ship

AI-generated patches in 45 min. Per-file rescan to verify fixes. 8 attack surfaces in one pipeline.

CO
Compliance officer

Map findings to controls automatically

43 frameworks from CIS to SOC 2. Evidence links, severity SLA tracking, structured audit logging.

DO
DevSecOps engineer

Automate security into every pipeline

Webhook triggers, SARIF export, Jira/Slack/Teams integration, CI/CD YAML templates. Zero manual gates.

SC
Startup CTO

Replace 3–4 point tools with one platform

SAST, SCA, secrets, IaC, cloud, network, runtime, and business logic — one scan, one report, one CI/CD integration.

SR
Security consultant

Deliver client reports faster

Multi-project scoping, per-client isolation, exportable SARIF, white-label ready.

Enterprise-grade from day one

Security, access, and compliance

CWE, OWASP, CVSS, NIST, PCI DSS — auto-tagged on every finding. Role-based access, encryption, audit logging, and 43 compliance frameworks built in.

Access Control

  • 5 built-in roles (Admin, SecOps, Engineer, Viewer, Custom)
  • 40+ granular permissions
  • Project-scoped access
  • API key scopes

Data Security

  • AES-256-GCM encryption at rest
  • TLS 1.3 in transit
  • Source code never leaves your environment
  • PostgreSQL row-level security per org

Deployment

  • Enterprise On-Prem or Krita Managed Infra
  • Air-gapped by default in both modes
  • External features opt-in only
  • Kubernetes or bare-metal deployment

Audit & Compliance

  • Structured JSON audit logging
  • 7-year retention
  • All changes tracked with before/after
  • IP and user agent logging
For engineers — every finding, already tagged
auth/login.ts · line 47CRITICAL
SQL Injection — unsanitised input into cursor.execute()
CVSS: AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H · exploit score 9.1/10

user_input flows from the HTTP request body directly into cursor.execute(query) without parameterisation or escaping. An attacker can exfiltrate the full database with a single crafted request.

Standards · auto-tagged on every finding
CWE-89OWASP A03:2021CVSS 9.8ASVS V5.3.4 · L2NIST SI-2PCI DSS 6.3MITRE T1190
Generated:
AI fixExploit scenarioJira ticketSLA timer · 4h
For compliance teams — what we cover and how
Requirement
What it demands
How we cover it
Req 6.2
Identify security vulnerabilities in all system components
SAST + SCA on every push — all languages, all dependencies
Req 6.3
Develop software resistant to attack
AI deep scan per file + cross-file business logic reasoning
Req 6.4
Address vulnerabilities by risk level with defined SLAs
4h CRITICAL · 24h HIGH · 72h MEDIUM — auto-enforced, blocks merge
Req 11.3
Authenticated internal vulnerability scanning
Scheduled scans with audit trail + SARIF evidence export

Ready to see what pattern matchers miss?

Request early access and run a scan inside your own environment. Your source code never leaves.

Source code never transmittedNo credit card required45 min to first finding
CWE·OWASP Top 10·CVSS v3.1·SOC 2·PCI-DSS·NIST 800-53