# Kolega.dev

Kolega.dev is an AI-powered code remediation engine for finding security vulnerabilities and generating merge-ready fixes.

## Primary Pages
- [Kolega.dev - The Code Remediation Engine](https://kolega.dev): Product homepage for Kolega.dev, an AI security remediation platform that scans repositories and creates merge-ready security fixes.
- [Documentation | Kolega.dev](https://kolega.dev/docs): User documentation for setup, integrations, platform workflow, scanning, remediation, and troubleshooting.
- [API Reference | Kolega.dev](https://kolega.dev/docs/api): Public API reference with authentication, request examples, endpoint definitions, and schemas.
- [Security Noise Tax Calculator - Calculate Your Team's Hidden Costs | Kolega.dev](https://kolega.dev/noise-tax-calculator): Interactive calculator for estimating engineering time and budget lost to security false positives.

## Content Sections
- [Security Wins | Kolega.dev](https://kolega.dev/security-wins): Security research writeups showing vulnerabilities found and fixed in public repositories.
- [Security Advisories | Kolega.dev](https://kolega.dev/advisories): Security advisories and vulnerability reports from Kolega.dev research.
- [Blog | Kolega.dev](https://kolega.dev/blog): Application security articles, product education, and industry analysis.

## Recent Articles
- [OpenClaw Security Assessment](https://kolega.dev/security-wins/openclaw-security-assessment): Identified 6 security vulnerabilities including command injection via unrestricted shell execution in elevated mode, path traversal in sandbox path resolution and media handling enabling arbitrary file read/write, insufficient rate limiting on authentication allowing brute force attacks, unsafe dynamic module loading, and sensitive data exposure logging third-party PII without redaction. 7 additional defence-in-depth recommendations were provided. All 13 reported items had PRs submitted for remediation.
- [Phase Security Assessment](https://kolega.dev/security-wins/phase-security-assessment): Identified 9 security vulnerabilities including critical authorization flaws from double-negative logic errors bypassing permission checks, type errors in authorization function calls, missing IDOR checks on credential retrieval and lease operations, insecure token transmission in GitLab OAuth adapter, and cross-organization payment method deletion. All 9 reported items were remediated across multiple PRs (#722-731).
- [Agenta Security Assessment](https://kolega.dev/security-wins/agenta-security-assessment): Identified 8 security vulnerabilities including 4 critical code execution flaws in the RestrictedPython sandbox implementation allowing arbitrary system command execution, 2 SSRF vulnerabilities in testset import and webhook evaluator endpoints, and authorization logic bugs enabling unauthorized modification of testsets. All 9 reported items were remediated and merged into release v0.77.1.
- [Cloudreve Security Assessment](https://kolega.dev/security-wins/cloudreve-security-assessment): 8 security vulnerabilities were identified across the platform, including a high-severity timing attack in HMAC signature validation. Other issues included plaintext database credentials appearing in logs, errors in lock token validation logic, insufficient salt entropy in password hashing, race conditions during node pool initialization, WOPI-related error information disclosure, and missing rate limiting on authentication endpoints. Five of these issues have since been remediated and merged into release 4.11.0.
- [vLLM Security Assessment](https://kolega.dev/security-wins/vllm-security-assessment): Identified 1 critical vulnerability involving unsafe torch.load() calls without weights_only verification in the tensorizer module, potentially allowing arbitrary code execution through malicious checkpoint files. Fix was remediated and merged into PR #32045.
- [Qdrant Security Assessment](https://kolega.dev/security-wins/qdrant-security-assessment): Identified 1 vulnerability involving unsafe memory access without complete bounds validation in the CSR loader, where get_unchecked() operations could lead to out-of-bounds memory reads from maliciously crafted CSR files. Fix was confirmed and merged into PR #7884 within 24 hours of disclosure.
- [Weaviate Security Assessment](https://kolega.dev/security-wins/weaviate-security-assessment): Identified 2 security vulnerabilities including unsafe AWS credential handling in the S3 backup module allowing attackers to inject credentials via HTTP headers to redirect backups to attacker-controlled buckets, and an SSRF vulnerability via unvalidated header-based URL override in the Anthropic module. Both issues confirmed and tracked in GitHub issue #10146 for resolution.
- [Langfuse Security Assessment](https://kolega.dev/security-wins/langfuse-security-assessment): Identified 4 security vulnerabilities including an SSRF vulnerability in the PostHog integration allowing requests to internal services and cloud metadata endpoints, cleartext storage of S3 credentials in the database, and missing server-side validation for data retention and member count limits (accepted as low-impact risks). Fixes merged in PRs #11311 and #11395.
- [NocoDB Security Assessment](https://kolega.dev/security-wins/nocodb-security-assessment): Identified 5 security vulnerabilities including 1 critical SQL injection in Oracle client, 2 high-severity SSRF issues in attachment uploads, WebSocket authentication bypass, and information disclosure. Delivered comprehensive fixes with tested pull requests, fixes have been implemented by NocoDB on their own internal private branches.
- [Agentic AI Framework Security: OpenClaw as a Case Study for Industry-Wide Standards](https://kolega.dev/advisories/agentic-ai-framework-security-openclaw-as-a-case-study-for-industry-wide): OpenClaw is the AI agent framework with 200,000+ GitHub stars and shell access to users' machines. It represents a critical gap in the agentic AI ecosystem: powerful tools shipping without the security practices their capabilities demand. This advisory documents our first-party findings (two CVSS 9.9 critical vulnerabilities in the OpenClaw codebase), pulls together the broader security crisis across agentic frameworks, and calls for the industry to adopt and enforce emerging standards before large-scale breaches become inevitable.
- [React2Shell Vulnerability Emergency](https://kolega.dev/advisories/react2shell-vulnerability-emergency): Critical RCE vulnerability in React 19 Server Components affecting 44% of global applications. Analysis of exploitation timeline, threat actors, and remediation strategies.
- [We built 24 apps with AI. Three platforms. 561 vulnerabilities.](https://kolega.dev/blog/we-built-24-apps-with-ai-three-platforms-561-vulnerabilities): Eight app categories. Three platforms. Same brief on every one. 561 vulnerabilities later, here's what the data says about vibe coding.
- [When SOC 2 Type II Compliant Means Anyone Can Read Your Source Code](https://kolega.dev/blog/when-soc-2-type-ii-compliant-means-anyone-can-read-your-source-code): Lovable spent 2025 turning security into a brand. Then a free account read everyone's chats, credentials, and code. Their response: That is intentional behaviour.
- [Your "Won't Fix" Backlog Just Became a Zero-Day Catalogue](https://kolega.dev/blog/your-wont-fix-backlog-just-became-a-zero-day-catalogue): Anthropic's Mythos weaponised the "Won't Fix" backlog. Here's why chain disruption and not faster patching is the only defence that scales.
- [Vibe Coding Is Safe for Production When You Do These 6 Things](https://kolega.dev/blog/vibe-coding-is-safe-for-production-when-you-do-these-6-things): Six things make vibe coding production-ready: secret scanning, dependency verification, API-level auth testing, semantic analysis for business logic, automated scanning on every push, and security headers. Here's exactly how to set each one up.
- [We Ran 15 Security Scanners Against Real Vulnerabilities. The Results Aren't Pretty](https://kolega.dev/blog/we-ran-15-security-scanners-against-real-vulnerabilities-the-results-arent): We tested 15 security scanners on 796 real vulnerabilities in 26 Python repos. SAST tools like Semgrep and Snyk caught less than 18%. LLMs like Claude Sonnet 4.6 caught 50%. Kolega.Dev caught 81%. All data, scoring code, and results are open-source at realvuln.kolega.dev.
- [The 8 Most Common Vulnerabilities in AI-Generated Code](https://kolega.dev/blog/the-8-most-common-vulnerabilities-in-ai-generated-code): 45% of AI-generated code fails security tests. These 8 vulnerability patterns keep shipping in vibe-coded apps - and your SAST tools miss most of them.
- [Your Security Scanner Was the Weapon: The TeamPCP Supply Chain Attack](https://kolega.dev/blog/your-security-scanner-was-the-weapon-the-teampcp-supply-chain-attack): On March 19, a routine Trivy scan stole your cloud credentials. Here's the full TeamPCP attack chain, why every detection tool missed it, and what to fix today.
- [Y Combinator Just Celebrated Building a Generation of Insecure Startups. Nobody Noticed.](https://kolega.dev/blog/y-combinator-just-celebrated-building-a-generation-of-insecure-startups): Demo Day celebrated 95% AI-generated codebases. But 45% of AI code fails security tests, and traditional SAST can't catch it. Here's the math they skipped.
- [We Needed a Benchmark That Didn't Exist. So We Built One.](https://kolega.dev/blog/why-we-built-our-own-security-benchmark): There’s a question nobody in security tooling answers cleanly: how do you know your tool would actually catch real-world vulnerabilities?
- [You Can't Secure OpenClaw. You Can Secure Yours.](https://kolega.dev/blog/you-cant-secure-openclaw-you-can-secure-yours): This started as a mission to fix agentic AI security. It ended somewhere I didn't expect.

## Policies
- [Privacy Policy | Kolega.dev](https://kolega.dev/privacy): Privacy policy for Kolega.dev, including code handling and AI processing commitments.
- [Terms of Service | Kolega.dev](https://kolega.dev/terms): Terms governing Kolega.dev accounts, subscriptions, acceptable use, and service usage.
- [Refund Policy | Kolega.dev](https://kolega.dev/refund): Refund and cancellation policy for Kolega.dev, including regional statutory withdrawal rights, waiver on meaningful use, auto-renewal terms, and the Paddle refund process.
- [Responsible Disclosure Policy | Kolega.dev](https://kolega.dev/responsible-disclosure): Responsible disclosure process, timelines, scope, and security contact information.

## Machine-Readable References
- [Sitemap](https://kolega.dev/sitemap.xml)
- [Robots policy](https://kolega.dev/robots.txt)
- [Docs-specific LLM guide](https://kolega.dev/docs/llms.txt): Focused index for documentation and API reference.
- [Public OpenAPI schema](https://kolega.dev/docs/api/openapi.json)

## Agent Guidance
- Prefer canonical URLs listed here and in sitemap.xml.
- Use the public API reference for endpoint details and schemas.
- Do not treat marketing examples as customer-specific security findings.
- For security disclosure questions, use the Responsible Disclosure Policy and security@kolega.ai.