The three layers
Everything in a Synclair clone is one of three things:Brain
The portable governance model — token vocabulary, the tiers, the docs contract,
search, the invention gate, the knowledge layer. Stable across projects.
Adapter
The platform-specific seams — how a preview is depicted, how tokens are exported,
how components are distributed. Swapped per target (web, React Native, SwiftUI).
Seed
Your project’s own content — brand ramps, theme, product name, knowledge, the
component catalog, the domain skill. Reseeded per project; never syncs.
What that means for modifying
The reseed is a file/directory operation, not untangling code. The seed lives in
known places (
lib/system/seed/, your theme, your knowledge, your catalog). Swap
those; keep the rest.Fit it to your setup
Common “can I make it fit my setup” questions, answered:- “Can I run it on a different port / mount path?” The hub mount point is a single swappable constant, and it runs on port 4100 by default — both are yours to change.
- “Can I target mobile, not web?” Yes — that’s what the adapter layer is for. The hub UI stays web/shadcn; only the adapter changes.
- “Can I add my own skills and agents?” Yes — that’s expected. Just clear the capability gate (prove nothing existing covers it) so the set stays coherent.
- “Can I restructure the hub?” You can, but that’s Brain — see the caution above.
Keeping the option to pull updates
Because your clone keeps the foundation as a gitupstream remote, you can choose
to pull foundation improvements later — as an ordinary git merge, on your terms.
Modifying seed freely costs you nothing here (seed conflicts resolve “keep
ours”). Heavily forking brain is what makes that merge harder — so if you find
yourself changing brain, ask whether it’s actually a foundation improvement worth
contributing upstream rather than a private fork. The synclair-sync skill handles
both directions.
Still have questions?
The fast-answer FAQ covers the rest.