The short answer
They are competitive in thesis, not identical in scope. Ent says it observes, investigates, and intervenes across human, AI, and application activity. ThumbGate is narrower: it evaluates an AI agent's proposed tool action and returns an inspectable ALLOW, WARN, or DENY before execution.
That narrower boundary is useful when the immediate problem is concrete: stop a force-push, secret leak, destructive shell command, unsafe dependency install, or repeated off-scope edit inside an agent workflow.