App Discovery Workflow

1. Use aws-mcp to call sagemaker:ListMlflowApps in the current region.

2. Filter results:

   • Show apps with status ACTIVE or CREATING
   • Exclude apps with status DELETING, FAILED, or DELETED

3. Rank results:

   • Apps with AccountDefaultStatus=Enabled appear first (account-wide default)
   • Then sort by most recently modified

4. Present to user:

   • Show: name, ARN, status, MLflow version, creation time
   • If exactly one ACTIVE app: suggest it directly — “You have one MLflow app: {name}. Use this one?”
   • If multiple ACTIVE apps: present a numbered list, ask user to choose
   • If zero apps found: inform user — “No MLflow apps found in {region}. Would you like me to create one?”

Edge Cases

• App in CREATING status: Inform user it’s still provisioning. Offer to wait (poll with aws-mcp sagemaker:DescribeMlflowApp until ACTIVE) or pick another app.
• All apps are FAILED: Suggest creating a new one rather than troubleshooting failed apps.
• Wrong region: If user expected to find an app, ask if they want to check a different region before offering to create.