The biggest constraint stems directly from OS-level permission models and ecosystem fragmentation. Because the engine relies on editing local host configurations, it is currently locked entirely to desktop environments, leaving mobile platforms completely unaddressed. If an operator locks down their desktop workstation but has their phone sitting unmonitored right next to the keyboard, the focus wall is easily broken.
Additionally, because the base Model Context Protocol architecture is inherently stateless, tracking continuous time-blocks or managing persistent session states across an unexpected system reboot or server restart introduces real continuity hurdles. There is also a risk of "panic blocking" errors; if your agent locks out a top-level domain that you unexpectedly need for an immediate dependency check or API documentation, you have to actively prompt the agent to release it—which can cause friction if the assistant is pre-configured to deny unblock requests until a specific project task is marked complete.
LockIn MCP
Thank you so much for the review! I'm actively working on mobile support in the near future!