Skip to content
Request beta access

Inspect the public surface, attach bounded repository evidence, and turn findings into a release decision.

Passive public checks only. No destructive tests.

No account yet? Request beta access

Built for projects using Cursor, Claude Code, Codex, GitHub Copilot, Windsurf, VS Code, Replit, Google Antigravity, Lovable, Bolt, v0, Base44, GitHub, Vercel, Supabase and Firebase through clearly labelled MCP, CLI, Git or URL paths.

Vybentis

Platform

Start with the public surface, add code context only when it is needed, then carry one evidence backed decision into the fix.

scan = await vybentis.scan(url)gate = scan.releaseGate()fix = gate.nextAction()
Learn more
Card 1 / 6
WHY VYBENTIS

Your agent writes fast.Vybentis checks what actually shipped.

01

Decisions that explain themselves

Every finding keeps the observed evidence, deterministic rule ID, confidence and severity that produced the release decision.

  • Deterministic checks, not model only opinions
  • Affected routes and surfaces stay attached
  • Clear evidence thresholds for every state
02

Context only when it earns its place

Start with the public URL. Attach one read only repository snapshot only when the release question needs code context.

  • Public and repository scope stay visibly separate
  • One resolved commit, read only
  • Secrets are redacted before persistence or explanation
03

A precise job for the coding agent

Turn one observed finding into a bounded fix prompt that preserves scope, evidence and the exact verification target.

  • Works through MCP, CLI and portable reports
  • Change target and verification rule stay attached
  • No vague AI remediation
04

GDPR checks with a clear boundary

Complete Scan observes known tracking requests and cookies in a fresh isolated browser profile before any consent interaction. Quick Scan excludes privacy checks.

  • Known analytics and advertising endpoints and cookie names are classified conservatively
  • Supported signals include Google Analytics, Google Ads, Meta Pixel, Microsoft Clarity, Hotjar, LinkedIn Insight and TikTok Pixel
  • Technical signals support review and are not a GDPR compliance verdict
Card 1 / 4

Release intelligence for coding agents

One scan turns your live app into a fix your agent can trust.

Vybentis checks the public surface, keeps the evidence attached and verifies the same rule after the change.

Passive only Evidence locked Same rule
Vybentis Release Workspace
connected
SCANSCN_8Q14
MODEPASSIVE
REQUESTS18
ELAPSED08.4S
Read the live appactive
Build the evidencequeued
Send and verifyqueued

Input

Live surface

live
GET / read only
https://yourapp.dev
Target connectedready

Record

Evidence attached

live
EVIDENCE STREAM immutable
Security headerswaiting
Search metadatawaiting
Public routeswaiting
Accessible nameswaiting

Output

Coding agent handoff

live
AGENT HANDOFF scoped
Waiting for changepending

Current stage: Connect. Live app.

npx vybentis scan https://yourapp.dev

Observe the real app

Read only checks collect evidence from the public surface.

Keep the rule attached

The coding agent gets the finding, its evidence and the expected result.

Close with proof

Vybentis runs the same rule again before the finding can close.

Card 1 / 3

For developers

Build with the Vybentis SDK.

Bring deterministic findings, evidence and verification into the tools you already use.

Explore the SDK
Contract

Evidence contract

Rule ID, observed evidence, confidence and severity stay attached from scan to report. Explanations cannot rewrite the finding.

Learn more
Verifier

Verified events

A rescan changes state only when the documented evidence threshold is observed. Verified describes that scoped check, not a guarantee.

Learn more
Card 1 / 2
Handoff

Context aware handoffs

handoff.ts
const apiBase = "https://api.vybentis.dev/v1";const findings = await fetch(  `${apiBase}/scans/${scanId}/findings`,  { headers: {    Authorization: `Bearer ${token}`,    "X-Vybentis-Organization-Id": organizationId  } }).then((response) => response.json()); const finding = findings.data.find(  ({ rule_id }) => rule_id === "SEC_HEADER_002");if (!finding) throw new Error("Finding not returned"); await mcp.callTool({  name: "vybentis_fix_prompt_get",  arguments: { finding_id: finding.id }});

Scope, observed evidence and the exact verification target travel together in one agent ready handoff.

Learn more
Portable

Model & framework agnostic

Output destinations shown include Claude, Mistral, ChatGPT, Gemini, GLM and Qwen. These portable destinations do not imply first party provider connections.

Export the same bounded evidence to the model or framework already in your workflow. The deterministic finding stays unchanged.

Learn more
Card 1 / 2

Application check

Review your application with Vybebat.

Vybebat reviews your code, logic and architecture before release.

Open Vybebat

Security

Security vulnerabilities and trust boundaries relevant to the agreed scope.

Example questions. We agree the exact scope with you.

Questions we investigate

  • Can one account access another account’s data?
  • Are permissions checked on the server, too?
  • Are credentials or tokens publicly exposed?

The outcome is a tailored set of findings and priorities. Remediation implementation is not included.

How it works

  1. Request

    Tell us what you are building, the live URL if public, and the question you want answered.

  2. Agree scope and access

    We confirm what to review and the safest way to share only the files needed through a read-only GitHub repository or agreed file handoff. Do not send credentials or secrets in your initial request.

  3. Review findings and priorities

    Receive a focused review with evidence, priorities and clear next decisions for your application.

This is a requested, tailored review, not an automated instant scan or a guarantee of complete security.

FAQ

Questions before you scan.

Clear answers about scope, safety, verification and the €19 Complete Scan.

Ask Vybentis
Scope, without fine print.
Direct answers about what Vybentis does and does not claim.

Scan your app. See what matters. Send one clear fix to your coding agent.

Start nowSee pricing