familiar-second-brain

Familiar Second Brain is a local server that lets AI assistants read and write notes to a Markdown folder on your computer. The folder acts as a shared memory store that multiple AI tools can access, so context you gather in one session does not get lost when the session ends or when you switch to a different tool.

The system has three parts. The first is a Markdown vault: a folder on your machine where all notes live as plain .md files. The second is an MCP server, which is a small background program that exposes the vault to any application that speaks the MCP protocol, a standard that AI coding tools like Claude Code, Codex, and Cursor support. The third is a skill for a desktop AI app called Kimi Work, which lets that app save session summaries and conversation transcripts directly into the vault.

Through the MCP server, connected tools can save a new note, search existing notes by keyword, read a specific file, or write to a file. Searches rank results by how well the keyword matches the title, tags, headings, links, and body text of each note, and return the surrounding context alongside the match.

Beyond the live MCP tools, the project includes several Python scripts you can run on demand. One generates a daily brief pulling from recent sessions, transcripts, and inbox notes. Another triages notes that landed in an inbox folder, suggesting where to file them, what tags to add, and what other notes to link. A third generates per-project memory briefs. A fourth audits the vault for duplicate notes and contradictions. A fifth resurfaces older notes that are due for review. All of these write their output back into the vault as plain Markdown files.

The project uses only Python's standard library, so there are no extra packages to install. Everything runs locally, and the repository deliberately excludes personal vault notes, API keys, and app databases.