Hister: a private full-content search index you host yourself

The Hister project presents itself as "your own search engine": it brings the pages you visit in a browser together with your local files into one private, full-content search index. The index, the stored page content and the indexing rules stay on the server you configure, and the software is free and licensed under AGPLv3, with its source published on GitHub.

How it works

Hister runs a simple loop of collect, index and find. A browser extension saves newly visited pages, while file watchers, browser history imports and a site crawler add other sources to the same index. The server extracts the parts that matter and indexes their full text rather than only titles and URLs. Search is available from the web interface, a terminal client, the command line and an HTTP API, and the same index can be reached by AI assistants through an MCP server.

Privacy by default

The main difference from hosted search services is where the data lives: the Hister server does not phone home or report what you search, requires no cloud service, and clients send indexed content only to the server you have configured. A complete personal setup can run on a single local machine — one binary starts a server with no configuration — while shared deployments support user-scoped access on SQLite or PostgreSQL and can be installed with Docker or Nix. Optional semantic search sends text only to the embeddings endpoint you specify, and browser extensions may request page favicons; you decide whether and where those connections run.

Search, versions and integrations

Full-text search supports field filters, quoted phrases, wildcards, negation, date ranges and user-defined query aliases. A clean stored preview opens beside the results, so you can see where a document came from without retracing your steps. Index rules cover skip and priority settings, document versioning that can retain earlier content, and sensitive-content checks. The project's own demo shows 1,924 indexed pages and six results returned from the local index in 0.04 seconds.