Search everything you've ever bookmarked.

ask it a question

Search the way you remember.

Ask it a question, "the talks on raft I saved last summer", and a local model first reads what the question says (the dates, the kind of page, what it points at) and searches with that. It shows what it understood, and --literal searches your words alone.

interests

See what you're into.

curio groups your library into topics and names them: areas, and the interests inside them. The map draws every page you've saved near the ones like it, so you can zoom from a whole area down to a single page. Names stay put as your library grows.

use it with claude

Answers grounded in what you've read.

One line, and Claude Code can search your library, open a saved page, find related pages, and see your interests. Claude Desktop and any other MCP client work too.

claude mcp add curio -- curio-mcp

• what have I saved about pricing experiments?
• find the post I kept on SQLite's WAL mode · and summarize it
• anything in my bookmarks on rate limiting? · compare them

everyday commands

Made for the terminal and the Web.

curio search "what you remember about it"   # search your library
curio status --follow                       # watch an import until it's done
curio ui                                    # open the dashboard in your browser
curio interests                             # the topics in your library
curio add https://example.com/article       # save one page
curio import safari                         # import more bookmarks later

To keep your Mac comfortable during a big import:

curio pause | resume            # stop starting new work, then pick up again
curio throttle gentle           # do less at once, so the fans stay quiet
curio schedule 22:00-07:00      # only work overnight
curio keep-awake on             # stay awake while importing, when plugged in

If something seems off, run curio doctor: it checks every part and says what to fix.

privacy

Your library, the search, and the AI models all live on your Mac.

curio runs its models through Ollama, on your Mac. There's no account and no telemetry. curio goes online to download the pages you bookmarked.

For pages that block it, curio can ask Jina Reader to fetch the page; that service sees the page's address. To turn this off, set fetcher.native.jina_fallback: false in ~/.curio/config.yaml.

get started

That's it.

$ brew install samsar/tap/curio
$ curio up

curio up walks you through the rest and asks before each step:

1. It checks that your Mac can run AI models locally.
2. It installs Ollama, the app that runs those models, if you don't have it yet.
3. It picks the right models for your Mac, tells you why, and downloads them.
4. It sets curio up to keep running in the background, even after a restart.
5. It asks where to import your bookmarks from: Chrome, Safari, Firefox, or an exported file. It shows how many are new and how long they'll take.

A few thousand bookmarks takes a couple of hours to fetch and index. curio tells you when to check back, and you can search while it works. Run curio up again whenever you like. If everything's fine, it says "Nothing to do: curio is up."

good to know

Questions

What does it need?

A Mac with Apple silicon (M1 or later), macOS 14 or newer, and Homebrew. 16 GB of memory or more is best, though 8 GB works. You'll also need about 5–25 GB of free disk, depending on your Mac.

What does it cost?

Nothing. curio is open source under the Apache License 2.0. The code, the design decisions and the roadmap are all on GitHub.

How do I upgrade?

Run brew upgrade curio, then curio up.

Can I start over?

curio up --fresh moves your current library aside to ~/.curio.bak-<date> and starts a new one. Nothing is deleted.

How do I uninstall it?

Run curio daemon uninstall, then brew uninstall curio. Your library stays in ~/.curio until you delete it.