I lose things I’ve already read. A rigging diagram, a Cloudflare config note, some half-remembered paragraph about batteries — I know I saw it, I just can’t get back to it. Hister indexes the full contents of pages you visit and files you keep, on a server you run, which is the shape of the problem I actually have.
The part I’d wire up first is the MCP client: full-text search over my own reading, exposed to an AI assistant, so it stops guessing and starts quoting things I’ve already vetted. Start local — single binary, ./hister listen, extension on, search a phrase from a page you just read. No config needed. I’d leave the optional semantic search off; that endpoint sends document text out, and the whole point is that it doesn’t leave.
The story — Hister is a private, self-hosted search engine in Go that full-text indexes the web pages you visit and local files you keep, searchable from a web UI, terminal, or an AI assistant over MCP. It ships as a single binary, with Homebrew, Docker, and Nix installs, browser extensions for Firefox and Chrome, multi-user support, no telemetry, and an AGPLv3 license. (Source)