wtf is memobsidian?

memobsidian connects your Obsidian notes vault and your Granola meeting notes into a single searchable memory that AI assistants like Claude Desktop, Cursor and Windsurf can query, and it keeps everything on your own computer instead of sending it to the cloud. It solves two specific problems: your AI assistant normally cannot see either your Obsidian notes or your Granola meeting notes, and Granola's free plan only keeps meeting notes for about 30 days before they disappear. Under the hood it runs a local search and memory service called Supermemory on your machine, paired with Ollama, a tool for running AI models locally, to turn your notes into a searchable index without any of it leaving your device. Once set up, an AI assistant connects to it through MCP, a standard way for AI tools to gain extra abilities, and can search or add to your notes using two commands called search_memory and add_memory. Getting started means having an existing Obsidian vault, then running a setup command that asks for your vault location and preferred Ollama model, asks whether you also use Granola, and then installs anything missing such as Ollama itself or the Supermemory service. After that, day to day commands let you sync your Granola notes into Obsidian, check that everything is running correctly, and start or stop the background service. The project is organized into separate pieces: a command line tool, an Obsidian plugin that pushes vault notes into the memory service, a Granola export tool, and the MCP server itself that your AI assistant talks to. It requires Node.js version 16 or newer, and optionally Homebrew on Mac to help install other dependencies. The README does not state a specific license type, only pointing to a LICENSE file in the repository.