Security
6 articles on security — from architectural deep-dives to production deployment guides.
The MCP Server Security Checklist: 30 Controls for Production Deployments
A practical, copy-paste security checklist for deploying MCP servers in production. Covers credential management, transport security, DLP, prompt injection defense, supply chain verification, and audit logging — with specific actions for each control.
What is an MCP Gateway? Enterprise Security for AI Agent Tool Execution
An MCP gateway is a security and governance layer that sits between AI agents and MCP servers, enforcing authentication, audit logging, DLP, and semantic intent classification on every tool call before it executes.
The Two-Gateway Architecture: Why Inference Proxies Cannot Protect MCP Tools
A monumental 2000-word deep dive into the 2026 'Two-Gateway' architecture. Learn the exact difference between an Inference Gateway (Cloudflare) and an MCP Execution Gateway (Vinkius Edge).
Why Your AI Agent Needs a Security Layer Before Touching Production Data — The CISO's Guide to MCP Governance
A technical guide for CISOs, CTOs, and security-conscious engineering leaders on why every MCP server connection to production data needs a governed security layer. Covers DLP, credential isolation, audit trails, context bleeding risks, and the enterprise governance framework that separates demo-grade AI from production-grade AI.
Context Bleeding: How JSON.stringify() in MCP Servers Leaks Databases
A formal CWE-200 vulnerability disclosure — with proof-of-concept code, CVSS scoring, ORM-level analysis, and CVE filing — targeting the architectural anti-pattern taught by AI SDK tutorials.
MCP API Key Management: From Plaintext JSON to Zero-Trust Credential Security
Every MCP client stores API keys in plaintext JSON files on developer laptops. This guide exposes the credential crisis in the MCP ecosystem and shows how Vinkius eliminates it with opaque tokens, encrypted vaults, SIEM streaming, emergency kill switches, and cryptographic audit trails.
Protect your AI agents with governed MCP
V8 sandbox isolation. SSRF protection. Cryptographic audit trail. Real-time kill switch.