Memtru helps developers test the reliability and security of AI agent memory. Run the Python CLI locally to check cross-user and cross-tenant isolation, stale information, deletion visibility and deleted data reappearing after background work. Connect your own system through an adapter in an authorized sandbox with synthetic data, review JSON and HTML evidence, then rerun after a fix. The public demo shows a deliberately vulnerable application failing and a corrected version passing. Results cover the scenarios and retrieval paths tested, not absolute security or physical data erasure.