Work directly in the GUI
Create “Personal Finance App,” add, move, and connect nodes yourself. Ask AI to Expand for scope or Deepen for detail, then accept, edit, or discard every suggestion.
How it works
Two paths turn ideas into actionable plans.
Organize the map in the human-controlled GUI or talk with an Agent. AI helps Expand and Deepen; people decide what becomes true.
01 / Choose a starting path
Talk with an Agent
Describe the problem. The Agent records each confirmed goal, feature, decision, task, and risk as its own node while you correct the map.
02 / Learn the map through one example
Hierarchy first: Tree
Put “Quick entry,” “Monthly budget,” and “Data export” under the app. That parent-child structure is the Tree.
Cross-links: Graph
Connect “Data export” to “Privacy risk,” or “Monthly budget” to “Notifications.” Cross-tree relations form the Graph.
Mainline and Branch
Make “manual entry first” the Mainline. Explore “bank import” in a Branch, then merge, keep, or archive it after validation.
Expand outward
Expand “Quick entry” into amount, category, date, notes, and validation. Suggestions remain reviewable.
Deepen inward
Deepen “Monthly budget” into done criteria, constraints, dependencies, risks, alternatives, acceptance checks, and todos.
03 / Technical words come last
Node types
The nine node types are Idea, Concept, Module, Task, Question, Decision, Risk, Resource, and Note. Expand, Deepen, Mainline, and Branch retain these exact contract terms.
Data and controls
Projects use local SQLite, full JSON backup, and readable Markdown export. After selecting a node in the desktop app, E runs Expand and D runs Deepen.
AI data flow
The project DB stays local. Expand and Deepen send only the selected node and context needed for the request to the model provider configured by the user; this website receives none of it. The provider may process it under its policy. Agent Port neither executes models nor stores agent API keys. See Security.