MCP Gateway Security: Lessons From Two Gateway CVEs

Two 2026 vulnerabilities in MCP gateway projects illustrate the security risks of weak control-plane authorization: CVE-2026-109492, rated CVSS 8.8, affected Microsoft’s open-source MCP Gateway by allowing authenticated users who could create adapters or tools to exploit user-chosen names, collide with Kubernetes resources, and potentially access, alter, delete, or redirect traffic to another tenant’s workload or the built-in tool gateway; version 1.0.0 fixes the issue through reserved names, cross-type uniqueness checks, ownership verification, and safer object handling. CVE-2026-125926 affected mcp-gateway-registry versions 1.0.3 through 1.34.0, allowing authorized server registrants or editors to configure upstream addresses targeting internal services, creating an SSRF risk; version 1.35.0 adds validation, while operators are advised to review existing configurations and logs. The discussion emphasizes that MCP security requires both control-plane protections for creating, modifying, and owning tools, adapters, servers, and infrastructure resources, and data-plane enforcement for each individual tool invocation. Recommended practices include using internally generated resource identities, tenant isolation, least-privilege service accounts, restricted network egress, separate permissions for management operations, validation of upstream destinations, secure credential handling, default-deny policies for high-risk tools, and comprehensive audit logs for configuration changes and tool-call decisions.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.