CybeDefend

We use analytics and ad-measurement cookies (Google, Meta). Nothing is sold, and you can decline.

DeclineAccept

Your coding agents' security, in one command.

Rules followed, dangerous actions blocked, code fixed before the commit, and every agent visible to your team.

$npx -y @cybedefend/vibedefend@latest installCopy

Send to my computer

Book a demo 5/5 on

EU or US region, chosen at install. We keep none of your code.

Claude Code

Cursor

OpenAI Codex

GitHub Copilot

Windsurf

One command. Five agents. The installer wires Claude Code, Cursor, Windsurf, GitHub Copilot and OpenAI Codex on the developer's machine, in one run.

Rules broken per ticket, same agent, same tasks.

• No tool2.28
• A rules file, kept by hand2.27
• VibeDefend0.28

Controlled study, 24 August 2026. Read the study
A rules file against VibeDefend

2. Your rules, served before each edit.

   Business rules learned from your code and security rules for your stack, picked for the file it edits.

   Claude Code is writing billing/refund.ts

   async function refund(order, amount) { if (amount > 500) await approvals.second(order);  await payments.refund(order, amount);  return receipt(order); }
   

   Refunds above €500 need a second approver.Approved once by Sofia M.

   Written with your rule, before the edit lands.

3. Dangerous commands never run.

   Schema drops, sudo, secret reads and force pushes are refused; the agent is told why and takes another way.

   OpenAI Codex wants to run

   rm -rf ./storage/uploads

   1. Read
   2. Checked against your guards
   3. Refused
   4. Run

   Blocked: protected path, user uploads

4. Fixed before the commit.

   When the agent finishes, its changes are scanned and it fixes each finding. Your project's findings are in its session.

   SQL injection, reachableCWE-89, src/db/users.repo.ts:42 "SELECT * FROM users WHERE id = " + id "SELECT * FROM users WHERE id = $1", [id]

   Found in the agent's diffFixed by the agent before the commit

5. Then proved, not assumed.

   After the commit, BLSA checks its business logic and proves each verdict: the rule, the path, the line.

   1. Refunds above €500 need a second approver.

      BLSA, proof mode

   2. POST /refunds

      Entry point

   3. approvals.second(order)

      The rule's check

   4. payments.refund(order)

      Runs after the check

   Holds: approved before it is paid

6. Everything it does, on one map.

   Every rule served, every refusal and every fix, traced to its agent and its developer in the Brain.

   Cursor

   1. src/billing/cart-checkout.ts

      3 rules served

   2. DROP TABLE invoices;

      Blocked

   3. CWE-89, src/db/users.repo.ts:42

      Finding fixed

   Traced to Sofia M., in the Brain

Claude Code is writing api/invoices.tsEvery query is scoped to the tenant. rules/data-access.md

Cursor wants to runDROP TABLE invoices;

Scanned when the agent finishesSQL injection, reachable CWE-89, src/db/users.repo.ts:42Fixed by the agent before the commit

BLSA, proof modeRefunds above €500 need a second approver.

Every agent, live in the Brainrule servedaction blockedfinding fixed

Nothing to rewrite in review.

Nothing to restore.

Nothing to chase after the merge.

Nothing taken on trust.

No one to ask. It is all on the map.

Without the rule

Written with the rule

Cursor

Refused

invoices, untouched

Clean commit

POST /refunds

approvals.second(order)

payments.refund(order)

Holds: approved before it is paid

Claude Code

Cursor

OpenAI Codex

GitHub Copilot

Windsurf

• DevelopersOne command, in the agent you already use.
• Engineering leadsThe same rules in every agent, and the savings measured.
• Security teamsEvery agent action, visible and traced in the Brain.

$npx -y @cybedefend/vibedefend@latest installCopy

Send to my computer

• Setup guides
• Claude Code
• Cursor
• OpenAI Codex
• Windsurf
• GitHub Copilot

In detail

// GET /api/orders/:id const order = await db.order.findUnique({ -  where: { id: req.params.id }, }); return res.json(order);

// GET /api/orders/:id const order = await db.order.findUnique({ +  where: { id: req.params.id, tenantId: req.user.tenantId }, }); +if (!order) return res.status(404).end(); return res.json(order);

-router.post('/refund', async (req, res) => { -  await stripe.refunds.create({ -    payment_intent: req.body.intent, -    amount: req.body.amount,  });  return res.json({ ok: true }); });

+router.post('/refund', requireRole('finance'), async (req, res) => { +  const intent = await getOwnedIntent(req.user, req.body.intent); +  const amount = clampRefund(intent, req.body.amount); +  await stripe.refunds.create({ payment_intent: intent.id, amount });  return res.json({ ok: true }); });

app.post('/webhook/stripe', async (req, res) => {  await orders.markPaid(req.body.id); -  await ledger.append({ type: 'charge', ... });  res.send('ok'); });

app.post('/webhook/stripe', async (req, res) => { +  const key = req.headers['idempotency-key']; +  if (await seen.has(key)) return res.send('dup');  await orders.markPaid(req.body.id); +  await ledger.append({ type: 'charge', key, ... }); +  await seen.add(key, '24h');  res.send('ok'); });

-logger.info('signup', { user: req.body });

+logger.info('signup', { +  userId: hash(req.body.email), +  region: req.body.region, +});

-// public.routes.ts -router.get('/admin/users', listAllUsers);

+// admin.routes.ts (auth + rbac applied at mount) +router.use(requireSession, requireRole('admin')); +router.get('/users', listAllUsers);

-if (flags.allow_anyone_to_export ?? true) {  return exportAll(); }

+if (flags.allow_anyone_to_export === true && req.user.role === 'admin') { +  audit.log('export.all', req.user);  return exportAll(); }

-router.post('/login', loginHandler);

+router.post('/login', +  rateLimit({ windowMs: 60_000, max: 8, keyGenerator: ipAndEmail }), +  loginHandler);

-router.get('/admin/export', requireAdmin, async (_req, res) => { -  const rows = await db.user.findMany({});  res.csv(rows); });

+router.get('/admin/export', requireAdmin, async (req, res) => { +  const rows = await db.user.findMany({ +    where: { tenantId: req.user.tenantId },  }); +  audit.log('export', req.user);  res.csv(rows); });

const total = new Decimal128(cart.sum); const rows = await db.query( "SELECT * FROM carts WHERE id = $1", [id] );

Before you install.

In CybeDefend's controlled study of 24 August 2026, the same agent applied 89% of precise rules with VibeDefend, against 12% with no tool and 13% with a rules file kept by hand.

Read the documentation The study, on GitHub

{"@context":"https://schema.org","@type":"Organization","@id":"https://www.cybedefend.com/#organization","name":"CybeDefend","legalName":"CybeDefend SAS","alternateName":["Cybedefend","Cybe Defend","Cybe-Defend","cybedefend"],"url":"https://www.cybedefend.com","logo":{"@type":"ImageObject","url":"https://www.cybedefend.com/brand/cybedefend-icon-512.png","width":"512","height":"512","caption":"CybeDefend logo"},"image":"https://www.cybedefend.com/brand/cybedefend-icon-512.png","description":"CybeDefend gives coding agents the context they lack. VibeDefend, its agent layer, serves your business and security rules to Claude Code, Cursor, Windsurf, GitHub Copilot and OpenAI Codex at the line they edit, stops dangerous commands before they run in Claude Code, Cursor and OpenAI Codex, checks every change before the commit and proposes new rules from your code for a person to approve. The Brain shows it live: which agents run in your organisation, who drives them, what they were served and what was stopped. CybeDefend Platform, open source, runs your scanners, removes duplicate findings and ranks each one by reachability.","slogan":"Intelligence is solved. Context isn't.","keywords":"context for AI coding agents, AI coding agent security, business rules for coding agents, guardrails for coding agents, VibeDefend, the Brain, agent-time security, Claude Code, Cursor, Windsurf, GitHub Copilot, OpenAI Codex, MCP, Model Context Protocol, open-source code security platform, SAST, SCA, IaC, container security, CI/CD security, secret detection, business logic, DevSecOps, Application Security","email":"contact@cybedefend.com","foundingDate":"2025-01","founder":[{"@type":"Person","@id":"https://www.cybedefend.com/#person-julien-zammit","name":"Julien Zammit"},{"@type":"Person","@id":"https://www.cybedefend.com/#person-florentin-ledy","name":"Florentin Ledy"},{"@type":"Person","@id":"https://www.cybedefend.com/#person-axel-paulin","name":"Axel Paulin"}],"member":[{"@type":"Person","@id":"https://www.cybedefend.com/#person-jason-lee","name":"Jason Lee","jobTitle":"Board Advisor"}],"address":{"@type":"PostalAddress","streetAddress":"177 Allée Clémentine Deman","addressLocality":"Lille","postalCode":"59000","addressCountry":"FR"},"location":[{"@type":"Place","name":"CybeDefend HQ, Lille","address":{"@type":"PostalAddress","streetAddress":"177 Allée Clémentine Deman","addressLocality":"Lille","postalCode":"59000","addressCountry":"FR"}},{"@type":"Place","name":"CybeDefend US, Seattle","address":{"@type":"PostalAddress","addressLocality":"Seattle","addressRegion":"WA","addressCountry":"US"}}],"sameAs":["https://www.linkedin.com/company/cybedefend","https://github.com/orgs/CybeDefend","https://x.com/cybedefend","https://discord.gg/UPzm4n6SAJ","https://www.instagram.com/cybedefend","https://www.facebook.com/people/CybeDefend/61591518974126/","https://www.tiktok.com/@cybedefend","https://trust.inc/cybedefend-2","https://www.youtube.com/@CybeDefend","https://www.g2.com/sellers/cybedefend","https://www.producthunt.com/products/cybedefend"],"knowsAbout":["Context for AI coding agents","AI coding agent security","Business rules for coding agents","Application Security","AI-assisted code generation","Static Application Security Testing (SAST)","Software Composition Analysis (SCA)","Infrastructure as Code security","Container and image security","CI/CD pipeline security","Secret detection and management","Model Context Protocol (MCP)","DevSecOps","Business logic security"],"areaServed":[{"@type":"AdministrativeArea","name":"European Union"},{"@type":"Country","name":"United States"},{"@type":"Place","name":"Worldwide"}],"contactPoint":[{"@type":"ContactPoint","email":"contact@cybedefend.com","contactType":"customer support","availableLanguage":["English","French","Spanish","Portuguese","Italian","German"],"areaServed":["EU","US","Worldwide"]},{"@type":"ContactPoint","email":"security@cybedefend.com","contactType":"security","availableLanguage":["English","French"]},{"@type":"ContactPoint","email":"press@cybedefend.com","contactType":"press inquiry","availableLanguage":["English","French"]}],"hasOfferCatalog":{"@type":"OfferCatalog","name":"CybeDefend products and services","itemListElement":[{"@type":"Offer","itemOffered":{"@type":"SoftwareApplication","@id":"https://www.cybedefend.com/#vibedefend","name":"VibeDefend","url":"https://www.cybedefend.com/en/vibedefend","description":"Your rules, served to every coding agent: business and security rules at the line the agent edits, dangerous commands stopped before they run in Claude Code, Cursor and OpenAI Codex, every change checked before the commit. Installs in Claude Code, Cursor, Windsurf, GitHub Copilot and OpenAI Codex."}},{"@type":"Offer","itemOffered":{"@type":"Service","name":"AI-BOM","url":"https://www.cybedefend.com/en/ai-bom","description":"AI component inventory aligned with EU AI Act and NIST AI RMF. Discovers every model, prompt path and AI library shipped in your codebase."}},{"@type":"Offer","itemOffered":{"@type":"Service","name":"Cybe Security Champion","url":"https://www.cybedefend.com/en/cybe-security-champion","description":"AI security copilot in your IDE (VS Code, JetBrains, Zed) with knowledge-graph reachability context."}},{"@type":"Offer","itemOffered":{"@type":"Service","name":"Cybe AutoFix","url":"https://www.cybedefend.com/en/cybe-autofix","description":"Autonomous AI agent that generates verified vulnerability fixes and opens ready-to-merge pull requests."}},{"@type":"Offer","itemOffered":{"@type":"Service","name":"Cybe Analysis","url":"https://www.cybedefend.com/en/cybe-analysis","description":"Autonomous AI agent for vulnerability triage with reachability scoring and exploit-path ranking."}},{"@type":"Offer","itemOffered":{"@type":"Service","name":"CybeDefend Platform","url":"https://www.cybedefend.com/en/platform","description":"Open-source code security platform: runs your SARIF scanners, removes duplicate findings, ranks each one by reachability and keeps its history. Community is free on your own servers; Cloud is hosted by CybeDefend with its own scanners, AI triage and AutoFix."}}]}}
{"@context":"https://schema.org","@type":"LocalBusiness","@id":"https://www.cybedefend.com/#local-business","name":"CybeDefend SAS","url":"https://www.cybedefend.com","email":"contact@cybedefend.com","image":"https://www.cybedefend.com/brand/cybedefend-icon-512.png","description":"CybeDefend gives coding agents the context they lack. VibeDefend, its agent layer, serves your business and security rules to Claude Code, Cursor, Windsurf, GitHub Copilot and OpenAI Codex at the line they edit, stops dangerous commands before they run in Claude Code, Cursor and OpenAI Codex, checks every change before the commit and proposes new rules from your code for a person to approve. The Brain shows it live: which agents run in your organisation, who drives them, what they were served and what was stopped. CybeDefend Platform, open source, runs your scanners, removes duplicate findings and ranks each one by reachability.","address":{"@type":"PostalAddress","streetAddress":"177 Allée Clémentine Deman","addressLocality":"Lille","postalCode":"59000","addressCountry":"FR","addressRegion":"Hauts-de-France"},"geo":{"@type":"GeoCoordinates","latitude":50.6293,"longitude":3.0573},"parentOrganization":{"@id":"https://www.cybedefend.com/#organization"},"sameAs":["https://www.linkedin.com/company/cybedefend","https://github.com/orgs/CybeDefend","https://x.com/cybedefend","https://discord.gg/UPzm4n6SAJ","https://www.instagram.com/cybedefend","https://www.facebook.com/people/CybeDefend/61591518974126/","https://www.tiktok.com/@cybedefend","https://trust.inc/cybedefend-2","https://www.youtube.com/@CybeDefend","https://www.g2.com/sellers/cybedefend","https://www.producthunt.com/products/cybedefend"],"openingHoursSpecification":{"@type":"OpeningHoursSpecification","dayOfWeek":["Monday","Tuesday","Wednesday","Thursday","Friday"],"opens":"09:00","closes":"19:00"}}
{"@context":"https://schema.org","@type":"WebSite","@id":"https://www.cybedefend.com/#website","url":"https://www.cybedefend.com/en","name":"CybeDefend","alternateName":"Business context and guardrails for AI coding agents","inLanguage":"en-US","description":"CybeDefend gives coding agents the context they lack. VibeDefend, its agent layer, serves your business and security rules to Claude Code, Cursor, Windsurf, GitHub Copilot and OpenAI Codex at the line they edit, stops dangerous commands before they run in Claude Code, Cursor and OpenAI Codex, checks every change before the commit and proposes new rules from your code for a person to approve. The Brain shows it live: which agents run in your organisation, who drives them, what they were served and what was stopped. CybeDefend Platform, open source, runs your scanners, removes duplicate findings and ranks each one by reachability.","keywords":"context for AI coding agents, AI coding agent security, business rules for coding agents, guardrails for coding agents, VibeDefend, the Brain, agent-time security, Claude Code, Cursor, Windsurf, GitHub Copilot, OpenAI Codex, MCP, Model Context Protocol, open-source code security platform, SAST, SCA, IaC, container security, CI/CD security, secret detection, business logic, DevSecOps, Application Security","publisher":{"@id":"https://www.cybedefend.com/#organization"},"copyrightHolder":{"@id":"https://www.cybedefend.com/#organization"},"copyrightYear":2026,"isAccessibleForFree":true}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"CybeDefend","item":"https://www.cybedefend.com/en"},{"@type":"ListItem","position":2,"name":"VibeDefend","item":"https://www.cybedefend.com/en/vibedefend"}]}
{"@context":"https://schema.org","@type":"SoftwareApplication","@id":"https://www.cybedefend.com/#vibedefend","mainEntityOfPage":"https://www.cybedefend.com/en/vibedefend","name":"VibeDefend","alternateName":"Repeat","applicationCategory":"SecurityApplication","applicationSubCategory":"DeveloperSecurity","operatingSystem":"CLI, MCP server (stdio and HTTP), Claude Code, Cursor, Windsurf, Antigravity, Gemini CLI, GitHub Copilot, Cline, Continue, Zed","description":"One command puts VibeDefend in Claude Code, Cursor, Windsurf, Copilot and Codex: your business and security rules at every edit, dangerous commands stopped.","url":"https://www.cybedefend.com/en/vibedefend","image":"https://www.cybedefend.com/brand/cybedefend-icon-512.png","keywords":"VibeDefend, AI coding agent governance, MCP server, Claude Code MCP, Cursor MCP, Codex MCP, Windsurf MCP, VS Code Copilot MCP, business rules, security rules, action guards, AI guardrails, agent audit trail","softwareVersion":"2026.05","author":{"@type":"Organization","@id":"https://www.cybedefend.com/#organization"},"publisher":{"@type":"Organization","@id":"https://www.cybedefend.com/#organization"},"offers":{"@type":"Offer","price":"0","priceCurrency":"USD","url":"https://www.cybedefend.com/en/pricing","availability":"https://schema.org/InStock","eligibleRegion":[{"@type":"Place","name":"European Union"},{"@type":"Place","name":"United States"}]},"featureList":["Learns the business-logic rules your codebase already enforces on the first scan","Serves the business and security rules that apply to the line the agent is about to edit","Continuous rule learning every few prompts, autopilot or manual","Memory that holds across long sessions and conversation resets","Native on Claude Code, Cursor, Windsurf, GitHub Copilot, OpenAI Codex, Gemini, Cline, Continue, Zed and Antigravity","EU and US data residency with per-tenant isolation","Audit trail signed with the agent identity for every decision","Deep installer for five agents, open MCP protocol for every other compatible client"]}
{"@context":"https://schema.org","@type":"Service","name":"VibeDefend - CybeDefend","description":"One command puts VibeDefend in Claude Code, Cursor, Windsurf, Copilot and Codex: your business and security rules at every edit, dangerous commands stopped.","url":"https://www.cybedefend.com/en/vibedefend","serviceType":"Agent-time code security, business and security rules served to coding agents, rule learning from your code","provider":{"@id":"https://www.cybedefend.com/#organization"},"areaServed":[{"@type":"AdministrativeArea","name":"European Union"},{"@type":"Country","name":"United States"},{"@type":"Place","name":"Worldwide"}],"audience":{"@type":"Audience","audienceType":"Software developers, AppSec engineers, platform teams, CISOs"},"category":"Cybersecurity / AI code security","isRelatedTo":{"@id":"https://www.cybedefend.com/#vibedefend"},"offers":{"@type":"Offer","url":"https://www.cybedefend.com/pricing","availability":"https://schema.org/InStock"},"keywords":"VibeDefend, AI coding agent governance, MCP server, Claude Code MCP, Cursor MCP, Codex MCP, Windsurf MCP, VS Code Copilot MCP, business rules, security rules, action guards, AI guardrails, agent audit trail"}
{"@context":"https://schema.org","@type":"FAQPage","mainEntity":[{"@type":"Question","name":"Does my code leave my machine?","acceptedAnswer":{"@type":"Answer","text":"What you send for analysis does, and it stays in the region you picked at install: Paris for EU customers on SecNumCloud-qualified infrastructure, Iowa for US customers, on our own self-hosted models. That means the first lines the agent is writing and the changes it made, analysed in transit: we keep none of your code, only the trail of the rules served and the actions refused. Only one check goes through a third-party AI API: Action Guard's, when the agent tries a dangerous action, through OpenRouter, under a contract that rules out any training on that data. The guard decides on your machine, findings never cross regions, and your code is never used to train a model."}},{"@type":"Question","name":"Which agents does it work with?","acceptedAnswer":{"@type":"Answer","text":"The installer wires Claude Code, Cursor, Windsurf, GitHub Copilot and Codex. Any other agent that speaks MCP works in bring-your-own-config mode, including Gemini CLI, Cline, Continue.dev and Zed."}},{"@type":"Question","name":"What does the free plan include, and what happens after?","acceptedAnswer":{"@type":"Answer","text":"Sign in and run the install command: your rules reach Claude Code, Cursor, Windsurf, GitHub Copilot and Codex, and the guards run in Claude Code, Cursor and Codex, with no card. To go further, we run a pilot with your team, then set a price with you, for your usage."}},{"@type":"Question","name":"Is this another scanner that floods the agent with alerts?","acceptedAnswer":{"@type":"Answer","text":"No. Legacy scanners run on commits and bury you in false positives. VibeDefend runs at agent-time on the diff being written, with reachability and framework awareness, and hands the agent one finding it can act on, with the fix."}},{"@type":"Question","name":"Can I read the study before I sign up?","acceptedAnswer":{"@type":"Answer","text":"Yes, and you should. The 36-page paper, the 90 transcripts, the blind audits and the six cases where the layer did not deliver are in the public repository linked on this page."}},{"@type":"Question","name":"What happens when VibeDefend blocks a command?","acceptedAnswer":{"@type":"Answer","text":"The command never runs. The agent receives the reason in its session, so it can take another way, and the refusal is logged for the team in the Brain."}},{"@type":"Question","name":"Where do the rules it serves come from?","acceptedAnswer":{"@type":"Answer","text":"From two places: rules proposed from your own code, which a person approves once, and the frameworks you choose, such as the OWASP Top 10, SOC 2 or GDPR, served as rules. Before each edit, VibeDefend hands the agent the ones that apply to the file it is changing."}},{"@type":"Question","name":"What is checked before the commit?","acceptedAnswer":{"@type":"Answer","text":"The changes the agent just made. At the end of its response, VibeDefend reviews them and hands each finding back with its fix, so the agent rewrites the line before the commit. The findings of your SAST, SCA, secrets, IaC and CI/CD scans are in the session too, live."}},{"@type":"Question","name":"Can the team see what each agent did?","acceptedAnswer":{"@type":"Answer","text":"Yes. Every rule served, every refusal and every finding is traced to its agent and its developer in the Brain, the map of your agents, so a lead or a security team reads it without asking anyone."}},{"@type":"Question","name":"My agent already checks its own code. Why add CybeDefend?","acceptedAnswer":{"@type":"Answer","text":"Each agent checks its own work, with its own settings. CybeDefend gives all five agents the same rules, your business rules included, served before the edit rather than reviewed after it, and keeps one record of what each agent was served, what was blocked and what was fixed."}}]}