Claude-mem is an open-source project on GitHub that provides persistent context memory across AI agent sessions, capturing and compressing session data for future use. It supports multiple AI tools including Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, and OpenCode. The project aims to maintain continuity of knowledge even after sessions end or reconnect, enhancing AI agent performance over time.
The system works by automatically capturing observations of tool usage during sessions, generating semantic summaries, and injecting relevant context back into subsequent sessions. Installation is straightforward, with support for various bots like Grok Bot via simple commands. Claude-mem offers both hosted memory and optional local observer modes, allowing flexibility depending on user needs.
Persistent memory in AI agents addresses a key limitation in current models that lose context between sessions. By preserving and compressing session data, claude-mem enables agents to retain project knowledge, improving efficiency and reducing repetitive input. This approach aligns with broader trends in AI development focused on long-term memory and contextual awareness, similar to efforts seen in other agent frameworks and tools.
Claude-mem is actively maintained on GitHub, where users can access documentation, installation instructions, and community support. The project’s compatibility with multiple AI platforms and its open-source nature make it a notable resource for developers seeking enhanced AI session continuity.