AtMem functions as a host-neutral Agent Black Box, offering a reversible memory control plane for AI agents. It ensures local-first, auditable memory by recording exact host-observed text and multimodal boundaries while limiting claims to agent-supplied data. This system empowers agents built with OpenClaw, Pydantic AI, LangChain, or LangGraph with governed long-term memory, context injection, and verifiable execution evidence, all managed and authorized locally by its private intelligence companion, AtBot.
主要功能
01Verifiable execution evidence and provenance tracking for agent operations.
02Governed retrieval, context injection, correction, and deletion of agent memories.
03Managed Governed Task State for work-in-progress, lifecycle, and authorization for agent tasks.
04Local-first, auditable memory for AI agents without external API dependencies.
05Native integration with OpenClaw, Pydantic AI, and LangChain/LangGraph frameworks.
0612 GitHub stars
使用案例
01Equipping AI agents with auditable, long-term memory for consistent performance and decision-making.
02Implementing a 'black box' for AI agent operations to ensure accountability, reviewability, and compliance.
03Governing and tracking the progress, blockers, and outcomes of AI agent tasks with revisioned authority.