wtf is cmem?

Claude-Mem gives AI coding assistants a persistent memory across sessions. Normally, when you end a conversation with an AI coding tool like Claude Code, it forgets everything that happened. This project fixes that by capturing what your agent does during each session, compressing it into summaries, and feeding relevant context back into future sessions automatically. The tool hooks into the lifecycle of your AI coding sessions. As you work, it observes what tools get used and what happens, then stores those observations in a local database. A background worker service processes everything and generates semantic summaries. When you start a new session, relevant memories from past sessions get injected so the AI has continuity about your project. It also provides search tools that let the AI query past history using natural language, with a layered approach designed to save tokens, first getting a compact index of results, then pulling full details only for the ones that matter. This is for developers and founders who use AI coding assistants regularly on ongoing projects and are tired of re-explaining context every time they start a new session. If you spent yesterday debugging an authentication bug with Claude Code, today's session can pick up where you left off without you re-describing the problem. It works with several coding agents including Claude Code, OpenCode, Codex, Gemini, Copilot, and others listed in the README. The project also supports optional real-time feeds to messaging platforms like Slack, Discord, or Telegram via the OpenClaw gateway integration. A few things stand out about how it's built. It uses a hybrid search system combining a local SQLite database with a vector database for semantic search, meaning it can find related context even when the wording doesn't match exactly. It also offers privacy controls, you can wrap sensitive content in <private> tags to exclude it from being stored. The README notes an available cloud sync feature for backing up memories, and configuration options let you switch between different workflow modes and languages. The project is Apache 2.0 licensed, chosen specifically so it can be embedded in various developer tools and production systems.