wtf is findandseek-engine?

FindandSeek Engine is a local-first tool that indexes the documents, code, and email sitting on your computer, then lets an AI assistant search through them using a protocol called MCP. Everything runs on your own machine using local AI models, with no cloud service involved and nothing ever sent off the device. The core idea is that each file gets understood once when it is first indexed, so later questions are cheap to answer. Instead of an AI assistant reading entire files every time you ask something, it gets back a short summary card with the key fact, the source filename, and a confidence level, and only pulls the full passage if that summary is not enough. The README reports that in their own testing over a large set of files, this approach used 56 percent fewer input tokens and cut costs roughly in half compared to an assistant that reads whole files for every question. The engine can read many common file types, including PDFs, Word documents, spreadsheets, presentations, emails, and even scanned documents or photos through optical character recognition. It requires either a Mac with Apple silicon, which runs the AI models directly on the chip, or a Linux machine running a tool called Ollama to run the models instead. Setup involves installing Python, running a setup command that downloads the needed model files, and starting a local server. Once running, a user tells the engine which folders to index, and it keeps watching those folders in the background, automatically re-indexing files as they change. The assistant connects to it as an MCP tool, gaining abilities like searching files, summarizing a document, looking up people or organizations mentioned in files, and drafting a folder cleanup plan, though the README stresses that no tool actually moves or deletes files on its own. Everything is also available as a plain web API for anyone who prefers not to use MCP. The full README is longer than what was shown.