Hister: a local, self-hosted search index of your browsing and files that agents query via MCP

Hister, an AGPLv3 self-hosted search engine, indexes pages you visit and files you keep, and exposes that index to AI assistants over MCP.

Hister describes itself as a "private search engine for the pages you visit and the files you keep." It indexes their full text so you can search from a web UI, a terminal, or an AI assistant connected through MCP, according to its GitHub repo. For builders, that means a local memory layer: your agent can search what you've already read instead of fetching the web again. It surfaced via @bibryam.

The README describes a short setup. Download the binary, run hister listen, open http://127.0.0.1:4433 and install the Firefox or Chrome extension. There's no telemetry and no mandatory cloud service. Homebrew, Docker and Nix installs are available. Semantic search is optional, and it sends document text to an embeddings endpoint you configure. The license is AGPLv3.

The project is young. A third-party deployment plan notes the repo was created 2026-01-04 and has shipped releases roughly monthly. That plan also notes there is no retention limit or disk quota. The v0.19.0 notes added ChatGPT and Hacker News extractors, Safari history import and url_re: search. Open issues point to rough edges. Near-duplicate URLs aren't merged (#744), and one proposal would stop semantic queries from waiting behind indexing (#801).

Some things are still unclear: the exact Claude Code or Claude Desktop config, the format of MCP responses, and parsing quality for PDF and Office files. A daily repo roundup also flags Office support as unconfirmed.

If you only need live access, there are alternatives:

Pick Hister if you want a searchable history of what you've read. Skip it if live browsing or file access is enough.

Sources