What is samemind?

Samemind is a memory system that lets you keep one consistent personal or project knowledge base across many different AI coding assistants, such as Claude Code, Cursor, Codex, and about a dozen others. The idea is that instead of each tool starting fresh with no memory of you or your project, you keep a single bundle of plain markdown files that any of these agents can read from and write to, so switching tools does not mean losing context. Everything lives in git and plain markdown files, with no database or background service required. A single command, samemind setup, detects which agent you are using, creates the memory bundle if you do not already have one, wires the necessary instructions into that agent's own configuration file, and sets it up as an MCP server so the agent can query it directly. Search works out of the box using a keyword-based method called BM25, and can optionally use a more advanced semantic search if you run a local embeddings server. Beyond plain notes, the bundle also holds an identity layer describing who the agent is and how it should behave, an ongoing work log recorded as an append-only ledger, and a simple kanban-style board for tracking in-progress work, blocked tasks, and future plans, all stored together rather than spread across separate tools. There is also a compatibility layer with Google's Open Knowledge Format, letting you export or import memory data in a shared, documented structure rather than a format tied only to this project. You can install samemind for a single project, or run it in a global mode that creates one personal bundle for your whole machine, usable across every project regardless of whether that project has its own local bundle. Setup steps are designed to be safely repeatable, so running them again does not create duplicate entries. The full README is longer than what was shown.