wtf is claude-mem?

Claude-Mem is a plugin for Claude Code, Anthropic's AI coding assistant, that gives it a persistent memory across separate work sessions. Normally when you close a Claude Code session and start a new one, the AI has no memory of what you worked on before. Claude-Mem solves this by automatically recording what Claude does during each session, compressing those observations into concise summaries using AI, and then injecting the most relevant pieces of that history back into future sessions so Claude has context about your project without you needing to re-explain everything. The system works by hooking into Claude Code's plugin system. A background worker process monitors each session, captures tool usage and observations, and stores them in a local database using ChromaDB, which is a vector database that allows semantic search. When a new session starts, the plugin retrieves the most relevant past observations based on what you are currently working on and surfaces them at the top of the context. There is also a web interface running locally that lets you browse your memory stream in real time, and a search skill that lets you explicitly query past work. You would use this if you work on long-running projects with Claude Code and find yourself constantly re-explaining project structure, past decisions, or ongoing tasks at the start of each session. It is also useful when switching between multiple projects, since the memory is project-scoped. The plugin supports Claude Code, Gemini CLI, and OpenCode. The tech stack is TypeScript running on Node.js 18 or later, with ChromaDB for vector storage and Claude's own agent SDK used to perform the AI-powered compression and summarization of observations. Installation is a single npx command.