This gives Claude a persistent memory system that stores everything in a single .claude/mind.mv2 file. It automatically captures your observations, decisions, and solutions as you code, then lets you search them with semantic queries like "What was the CORS solution?" The file is portable and git-friendly, so you can commit it or share it with teammates for instant context on a project. Memory classification happens automatically across types like discovery, bugfix, and refactor. The setup is straightforward: you run Node.js scripts to search, ask questions, or view timelines. It's honestly a solid approach to the context retention problem, especially if you work across multiple sessions or need to onboard people quickly without writing docs.
npx -y skills add memvid/claude-brain --skill mind --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
juliusbrussee/caveman
mattpocock/skills
shadcn/improve
obra/superpowers
forrestchang/andrej-karpathy-skills
vercel-labs/skills