Proactive discovery
Recommendations appear without requiring an account, peer, model, API key, or source configuration.
Open source · local first · alpha
Your personal AI node—local, open, and built to earn your trust.
RynMesh discovers useful public content, learns from feedback locally, verifies what it can, and keeps the owner in control.
Current source: v0.6.2
MIT licensed
No account required
v0.6.2Current release
MITOpen-source license
macOSTwo native architectures
P1Personal assistant milestone
Useful from first launch
No preference form between you and the first recommendation.
Ryn starts with a built-in public discovery catalog. Direction, topics, platforms, and feedback gradually reshape the local profile.
Local learning
More, Less, Hide, opens, topics, platforms, and written direction shape a profile stored by your node.
Inspectable evidence
Recommendation reasons, provenance, safety results, and source information remain visible instead of becoming a black box.
The released interface
Tell Ryn what deserves attention.
Preferences are optional. The owner can guide topics and platforms, inspect why each item appeared, and change the ranking through direct feedback.
Real v0.6-series product interface. Recommendations shown are broad starter choices before a profile has learned from the owner.
Open the product tourA node, not a cloud account
The local daemon is the authority.
Identity, preferences, feedback, and fetched content live beneath the owner’s RynMesh home. Network and model access are optional capabilities mediated by the node.
Read the architectureLocal-first
The personal-assistant experience works without joining a peer network.
Open implementation
The node, interface, protocol, registry, tools, and recommendation agent are MIT licensed.
Explicit boundaries
Cloud model access requires configuration; the web interface does not bypass the node.
Receipts over promises
Signed manifests, provenance, and safety evidence are designed to be inspected locally.
Build it with us
Choose accepted work. Reserve it. Make the evidence visible.
The contribution center reads the public GitHub backlog live. A reservation command prevents two primary contributors from unknowingly taking the same issue.