CVE Alert: CVE-2026-108714 – modelcontextprotocol – kotlin

MCP Kotlin SDK through 0.15.0 contains an uncontrolled memory allocation vulnerability that allows remote clients to exhaust server memory because Application.mcpWebSocket installs Ktor WebSockets without a maxFrameSize limit. Attackers can send small frame headers declaring payloads near 2 GiB over one or a few connections, forcing huge heap allocations and causing denial of service.

High operational risk: urgently identify exposed services and remediate, while confirming whether exploitation is active; KEV, SSVC, PoC and EPSS indicators were not supplied.

A remote attacker may be able to make an affected service unavailable without credentials or user action, disrupting applications and dependent workflows. The likely goal is service disruption or pressure on the organisation; there is no indicated direct route to data theft or system changes.

### Most likely attack path

An attacker connects to a reachable WebSocket endpoint and sends a crafted frame, requiring little effort and no victim interaction. The impact appears confined to the vulnerable service, so lateral movement is not indicated by the supplied scope data, though knock-on outages are possible.

### Who is most exposed

Prioritise internet-facing Kotlin services using the SDK’s WebSocket server extension, particularly MCP endpoints hosted in shared or business-critical environments.

Alert on abrupt heap growth, frequent full garbage collection, out-of-memory errors or process restarts.

Review WebSocket logs for unusually large declared frame lengths, repeated connections or abnormal disconnects.

Correlate endpoint availability and latency alerts with JVM memory and container restart events.

### Mitigation and prioritisation

Identify affected deployments and apply the vendor’s fixed release or guidance as soon as available; verify the fix in staging.

Until patched, restrict WebSocket access to trusted clients and enforce frame-size limits at the application or proxy layer.

Apply connection, request-rate and memory limits; monitor capacity and ensure recovery procedures are tested.

Check KEV, SSVC, PoC and EPSS status from trusted feeds before final queueing; escalate immediately if exploitation evidence emerges.

A considerable amount of time and effort goes into maintaining this website, creating backend automation and creating new features and content for you to make actionable intelligence decisions. Everyone that supports the site helps enable new functionality.

If you like the site, please support us on Patreon or Buy Me A Coffee using the buttons below.