Memtru

Memtru provides a robust framework for testing the memory integrity of software agents. Key features include: • Detect cross-user data leaks • Verify deletion visibility • Identify stale information retrieval • Reproduce failures with detailed reports • Compare test results before and after fixes This platform helps developers ensure that agents securely manage information, preventing sensitive data exposure between users or tenants. It addresses critical issues such as memory written for one user appearing for another, deleted information remaining retrievable, or old preferences outranking new ones. By simulating various data access scenarios, Memtru reveals where agent memory systems might fail to isolate data or correctly process updates and deletions. Memtru integrates directly into your existing testing environment via a Python CLI. Developers can run tests locally, generate comprehensive reports with evidence for each finding, and use these reports to pinpoint and resolve memory-related regressions. This makes it an indispensable tool for maintaining the reliability and security of data-driven applications. Designed for developers and engineering teams building sophisticated software agents, Memtru is perfect for validating data privacy, ensuring data consistency, and preventing regressions in how agents handle persistent information. It provides confidence that an agent's memory is trustworthy and behaves as expected across diverse operational contexts.