Start here

Building decision trees

A decision tree encodes your expert reasoning as branching questions, conditions, and conclusions. Once you Deploy it and set it Listed, your agent can call that logic over MCP and receive a structured report.

Next: Deploy, list, & delete → Connect your agent →

What a decision tree contains

Question node: A single fact or condition to establish. Has a set of valid answer options.
Conclusion node: An outcome or determination — the leaf of a branch. A decision tree can have multiple conclusions.
Edge: A connection from one node to the next. Carries the condition (answer) that must be true to follow that path.

Creating a decision tree

From your dashboard, click Create New Decision Tree. This opens the AI-assisted wizard. Describe your domain and paste in source material — procedure documents, reference guides, notes, or links. The wizard researches and drafts a full decision tree for you to edit.

You can also build or revise a draft through MCP chat. Your agent constructs the graph using SMEme design guidance; SMEme validates and saves it. See Build or revise a draft in chat.

Either way, the draft is a starting point. Always review and tune it in the editor before Deploy, and choose Listed or Hidden yourself.

The editor

The editor has two columns: a main view on the left and a sidebar on the right. The toolbar at the top has three tabs:

Tab	What it shows
Graph	Visual diagram of your decision tree. Nodes appear as cards connected by arrows. Click a node to select and edit it in the sidebar.
Checklist	A flat list of all nodes. Useful for reviewing question text and answer options across the whole decision tree without navigating the graph.
Tools	Deploy controls and the current deploy status (Live, Stale, or not yet deployed). See Deploy, list, & delete.

Sidebar

The sidebar is always visible on the right. When no node is selected it shows validation issues. Click a node in the Graph or Checklist and the sidebar loads that node's editor.

Adding nodes

Click Add node in the toolbar. Choose Question or Conclusion, fill in the text and answer options, and save.

Editing edges

Select a question node in Graph or Checklist view, then open the Connections section in the sidebar. There you can add, edit, or remove outgoing edges and set each edge's condition — which answer activates that path.

Saving and versioning

Changes are saved as you work. Saved changes do not automatically update what your AI tools run against — you need to Deploy for edits to take effect. See Deploy, list, & delete.

To remove a decision tree entirely, use the dashboard trash icon — see Deleting a decision tree.

Validation

While you edit, the sidebar lists validation errors and warnings for your saved graph. A badge in the toolbar shows the same counts. Errors must be fixed before deploy.

How to fix hints

Under many issues you'll see a How to fix: line — a short, concrete step. When an issue is tied to a node, click Jump to node to open it in the sidebar.

Common issues

Issue	What to do
Question must have options	Open the node and add answer options, then wire an edge from each choice.
Partial coverage / missing edge	One or more answer options have no outgoing path. Add a conditional edge for each option.
Unreachable node	Nothing in the graph leads to that node from the start. Connect it upstream or remove it if unused.
Edge to conclusion must be conditional	Set the edge condition to a specific answer instead of a default (catch-all).
Self-loop or cycle detected	Remove or retarget one edge in the loop so forward progress is possible.
Exactly one entry node	More than one node has no incoming edges. Connect extras from an upstream question, or delete them.