🌳 Hilka
Log how you think, not just what you decided.
A free, open-source decision-tree journal. Map a question as a tree of one-idea-per-card
thoughts — and the branches you reject stay forever, with the reason you killed them.
Build it by hand, or let your AI assistant build it for you.
Open the app → · Two ways to use · How it works
React 19 · TypeScript · Supabase · Cloudflare — Hilka (гілка) = branch.
Most notes capture conclusions. Hilka captures the reasoning — including the branches you
rejected, which is usually where the real thinking lives. Months later you re-read not just
what you chose, but why the alternatives didn't survive.
- 🪦 Rejected branches are first-class — dropping one demands a written reason; it stays on the canvas with that reason instead of vanishing.
- 🗺️ Two views of one tree — a pan/zoom canvas of cards on curved branches, or a linear chain feed.
- 🚦 A progress axis — every thought is
todo · doing · done · dropped.
- 📱 Phone & tablet ready — touch drag, pinch-zoom, bottom-sheet editing.
- 🎨 Two themes — Ink and Warm.
- 💾 Yours to keep — one-click Markdown / JSON export, free, open source.
Two ways to use Hilka
Hilka started as a canvas you build by hand. It now also lets any AI assistant build the
tree for you. Both are first-class and both are free — use whichever fits how you think.
✍️ Build it by hand
The original idea. Map a question as a tree of one-idea cards on a pan/zoom canvas (or a
linear chain feed). Branch into every option you're weighing; drag cards to reparent or
reorder. When you kill a branch, Hilka makes you write why before it lets you drop it — and
the dead end stays, greyed-out with its reason, so nothing is ever lost.
No setup, no account linking — just open it and start thinking.
→ Open the canvas and start your first chain.
|
🤖 Create it from your AI assistant
Mid-reasoning with Claude, ChatGPT, Codex, or Cursor, just say:
"make me a Hilka chain about whether to switch jobs" — or — "save this reasoning as a Hilka chain"
The assistant designs the tree — the branches, the dead-ends it ruled out and why — and
writes it straight into your Hilka account over a small MCP server.
Everything it creates is scoped to you.
→ Connect your assistant once, then create chains by asking.
|
Connect your assistant
Open your Hilka → Settings → Connect your AI assistant to see your personal server URL,
then pick one (most → least convenient):
- OAuth (no token): add the MCP server URL to your assistant and log in when prompted —
https://hilka-mcp.sviatik2408.workers.dev/mcp
- Claude plugin:
/plugin marketplace add Sviat838/Hilka- then /plugin install hilka@hilka
- Token (fallback): generate one in that screen and paste it as an
Authorization: Bearer header.
Architecture and setup of the MCP server live in worker/.
Why this matters (the longer version)
Hilka borrows from habits good decision-makers have used for decades:
- Decision journaling — write down a choice and its reasoning at the time, so you can
learn from it later instead of rewriting history.
- Issue / logic trees — break a fuzzy question into branches you can reason about one at
a time (the way consultants and engineers decompose problems).
- Mind mapping — think on a canvas, not in a wall of text.
This matters more, not less, in the age of AI:
You can outsource your thinking, but you can't outsource your understanding.
upstash
1270011
mike-coulbourn
obra