Eco mode

Eco mode was created to help teams run MCP Gateway more efficiently as AI infrastructure grows. In real deployments, idle MCP Gateway instances still consume Kubernetes cluster capacity, and over time that increases the cost of running AI workloads. Eco mode addresses this by automatically hibernating unused instances while preserving their state. So teams can optimize both cluster utilization and infrastructure cost without adding manual lifecycle management for users.

Eco mode hibernates idle MCP Gateway instances on a schedule to reduce Kubernetes resource usage. Hibernated instances keep their configuration, OAuth tokens, and persistent storage, and wake automatically the next time a user connects.

This is particularly useful for enterprises running multiple MCP Gateway instances that want to avoid the cost and overhead of keeping idle workloads online.

How it works

Eco mode has three parts:

1. Instance allowlist: Admins mark MCP Gateway instances as eco mode targets from the MCP Gateway Instances page.
2. Scheduled hibernation: A scheduled job automatically hibernates eligible instances based on a configured schedule.
3. Automatic wake-up: When a user reconnects, the system detects the hibernated state and automatically restores the instance before the system processes the request.

Eco mode pairs with hosted MCP servers, which uses the same instance lifecycle and phases.