A feed that answers to you
Real articles, papers, videos, podcasts and books from a built-in catalog, ranked on your machine without any model. More, Less and Hide change the ranking, and every pick explains why it appeared.
Open source · runs on your Mac · work in progress
Your own feed. Your own AI. A few friends, and no company in between.
Ryn is a desktop app that finds real content worth your time, learns from your feedback on your own disk, answers questions with a model running on your machine, and lets friends you invite share links and even lend each other their AI.
Current release: v0.7.0
MIT licensed
No account, no sign-up
What you get today
Install it, and it is useful in the first minute.
No account, no preference quiz, no API key. Everything on this list is in the current release and stays on your computer unless you decide otherwise.
Reading that stays yours
Bookmarks, progress, offline downloads, and one search across everything you have read, shared and asked. Pair a second computer and it keeps up, over an encrypted link you approve with a code.
Private AI on your own hardware
One click starts a bundled local model. Ask Ryn about the article you are reading, keep the history, switch models. Nothing leaves the machine unless you explicitly allow a cloud provider.
A small circle of friends
Invite a friend with a link or QR code, share an article, chat with attachments, follow what they choose to share, revoke at any time. A friend can let you use their AI, and they see what you ask it.
The real interface
Tell Ryn what deserves attention.
Preferences are optional. Guide topics and platforms, inspect why each item appeared, and change the ranking through direct feedback.
A real capture of the For You screen. The friends, Ask Ryn, offline reading and device screens added in v0.7.0 are not pictured yet.
Open the product tourWhere this is going
Four milestones, one at a time.
Each row shows what is already shipped and what comes next. Roadmap statements only become promises when they turn into accepted issues on GitHub.
Shipped
1 · Ryn Companion
Done
A desktop app with its own node. A feed with zero setup, local learning, a content viewer, reading history, offline reading and multi-device sync.
Next
Show per-source health and what the ranking has learned, with undo. Consolidate the daily digest. Windows and Linux builds.
In progress
2 · Trusted friends
Done
Invite by link or QR with endpoint review. Share, chat with attachments, follow friends' shares, revoke. Lend your local AI to a friend under your permissions.
Next
Rank friends' shares inside your own feed. Diagnose connections across a group of friends. A readable history of how reputation was earned.
Preview
3 · Working agent and services
Done
Private AI as the first service: encrypted node-to-node tasks, model switching, per-friend permissions, and a settlement ledger of non-transferable credits.
Next
Streamed answers. A general service protocol with metering and verified results. A budgeted agent that acts within limits you set, with a full audit trail.
Later
4 · Open network
Done
Signed identity, provenance and safety receipts, and a registry that coordinates without becoming an authority.
Next
Abuse resistance, trust weighting, moderation and appeals. Nobody you have not invited can reach your node until this work is finished.
Said plainly
What Ryn does not do yet.
We would rather you know before you download than find out after.
Read the v0.7.0 release notesNo money in it
Credits are non-transferable reputation. Nobody earns anything by running a node or lending their AI.
Friends have to install it too
Discovery of strangers is off by design. The sharing features need at least one friend on Ryn.
macOS only for now
Apple Silicon and Intel builds are signed but not notarized, so the first launch needs Control-click, then Open.
Rough edges
AI answers arrive whole rather than streamed, and friends' shares live in their own feed. Storage formats may still change before 1.0.
Open source, built in public
Engineers welcome. The backlog is open.
RynMesh is MIT licensed and developed by a very small team. If you want to work on local-first software, cryptographic pairing, recommendation systems, or a Python and React desktop app, every accepted issue comes with a written spec and a CI gate, and the maintainers review every pull request.
Start with the contributor setup, read the architecture, or ask in Discussions.