Documentation
Overview Building decision trees Plans & usage Deploy, list, & delete Download decision tree Connect your agent ChangelogDeploy, validate, list, & delete
Saving a decision tree in the editor does not make it available to your agent — you must Deploy (or Redeploy) it first. Then use Listed or Hidden to control MCP discoverability, watch Live vs Stale status, and use the trash icon to delete a decision tree permanently.
What Deploy does
Clicking Deploy does two things in order:
- 1 Runs readiness checks. A set of structural and logical checks on your saved graph. If any check fails, deployment stops and you'll see the issues to fix.
- 2 Builds and saves the deployed version. Your decision tree is packaged and stored as the live version your AI will use. This is what runs at evaluation time.
Readiness checks
These checks run automatically every time you deploy. Fix any failures before deploying.
| Check | What it verifies |
|---|---|
| Structure | At least one question node, one conclusion node, and a single root entry point. |
| Reachability | Every node can be reached from the root. No orphaned nodes. |
| Coverage | Every answer option on every question node has an outgoing edge. No dead ends except conclusion nodes. |
| Satisfiability | Every conclusion can be reached under at least one valid set of answers. No logically unreachable conclusions. |
| Node text | All nodes have non-empty text. All question nodes have at least two answer options. |
While you edit, the sidebar validation panel runs a lighter subset and shows How to fix hints under each issue.
Deploy status: Live and Stale
The Tools column on your dashboard compares your current saved graph with your last deployment. Your AI always evaluates against the deployed version — not unsaved or undeployed edits.
| Status | What it means |
|---|---|
| Live | Your saved graph matches the deployed version. Your AI is running the current version. |
| Stale | The graph has changed since the last deploy. Your AI is running the older version until you redeploy. |
| Deploy button | This decision tree has never been deployed. Click Deploy to make it available to agents for the first time. |
Why does my decision tree show Stale?
Any change to the saved graph after your last deploy — new nodes, removed nodes, changed edges, edited question text, or modified answer options. Stale is not an error; it means the deployed version no longer matches what you have saved.
What do I do?
Click Redeploy on that dashboard row. It updates the deployment from your current saved graph. The badge returns to Live after a successful redeploy.
Listed and Hidden
After a decision tree is deployed, the Listed toggle controls whether it appears in the tool list your MCP-connected AI can discover.
| State | Visible to AI? | Notes |
|---|---|---|
| Listed | Yes | Appears when your AI calls smeme_reasoning_list. You can invoke it by name. |
| Hidden | No | Deployed but hidden from your AI's tool list. Use for drafts or decision trees you're not ready to surface. |
Decision trees are Hidden by default after deploying. Toggle Listed on the dashboard row when you're ready.
Downloading a decision tree
Use the download icon in Actions to save the current decision tree as a JSON file. This is your editor decision tree — not the deployed version your AI uses. Download your decision tree has the full details.
Usage on the dashboard
The dashboard shows your plan and monthly meters: active decision trees, MCP tool calls, and AI-assisted wizard builds. The MCP (mo) column lists MCP usage per decision tree for the current UTC month (weighted by operation type). Meters turn amber at 80% of a cap; at 100% new decision trees, AI-assisted builds, or MCP tool calls are blocked until upgrade or reset. See Plans & usage for tier limits.
Deleting a decision tree
On the dashboard, the trash icon in Actions permanently removes a decision tree and every version in its family — deployed tools, evaluation history, and all. MCP-connected tools will no longer list or evaluate it. This cannot be undone and frees a decision tree slot.
Deletion is two steps: review what will be removed, then type
delete decision tree permanently
to confirm. In-progress decision trees (still generating) have a separate Delete button with a single confirmation.
Hidden is not delete. Hidden keeps the decision tree in your account and off your MCP tool list only.
To close your entire SMEme account (all decision trees, billing, and profile data), use Delete your account from Profile — not the dashboard trash icon.
See also
- Building decision trees — create and edit a decision tree before deploying
- Download your decision tree — export the saved graph as JSON
- Plans & usage — tier limits and quota behavior
- Delete your account — permanent account closure from Profile
- Connect your agent — add the MCP connector and invoke your listed decision trees