Start with a question.
Break a messy problem into separate parts, with no overlap and no blind spots.
Break a messy problem into separate parts, with no overlap and no blind spots.
Break a big question into separate parts that don't overlap (e.g. Grow revenue vs Decrease costs). Click a card for notes or + to split further.
A driver card is a lever you can pull; a Driver tree breaks down a calculation. Different branches can use different approaches via a badge or Details → Tree type.
Tell your agent your question, what you want to change and any constraints. It can choose a tree and build it with you.
Select the card where the question changes, then set its approach. That subtree uses it; deeper subtrees with their own approach keep theirs.
Why is profit falling? · Issue tree
├─ What changed in revenue?
│ └─ Revenue = orders × order value · Driver tree
│ ├─ Orders
│ └─ Average order value
└─ What changed in costs?
└─ Overtime may explain the rise · Hypothesis tree
└─ What do payroll records show?Fictional example. The driver branch explains a calculation; the hypothesis branch investigates a possible cause. Check each connection on its own terms. Selecting a method does not convert or validate the reasoning.
To use the wider tree’s approach again, select the subtree’s first card and choose Use parent type.
This type changed elsewhere. Review the latest setting first.
Changed elsewhere. Your draft is kept; review the latest saved version.
Paste the agent instructions into Claude Code, Codex, Cursor, or any agent - edits appear live on this screen.
Included for paid Wayfinder & Sprint members. Or connect your own agent above for free.
Answer in plain English. When ready, AI updates your tree directly on the canvas (Undo stays available).
Pick your goal and where things feel stuck. We lay out the main branches without overlap, attach your notes, and suggest 4 practical moves.
For standard browser chat without URL access: open with this tree, then paste the JSON reply back.