CARLA HQ
LIVE SIMULATOR

Try the Live Interactive Demo

Select a target column OfferDiscount and click Train & Analyze to simulate local WebGPU model training in Google Sheets.

CRM_Customer_Discount_Calibrated.csv
FileEditViewInsertFormatDataToolsExtensions
fx
ABCDEFGHI
CLVSegmentContractMonthsActivityScoreSupportTicketsIndustryAutoRenewalOfferDiscountCarla Pred
2HighEnterprise24851SaaSYesYes-
3MidSMB12405RetailNoNo-
4HighEnterprise36920FinTechYesYes-
5LowSMB6208HealthNoNo-
6MidEnterprise18653SaaSYesYes-
7LowSMB31512RetailNoNo-
8HighSMB24781SaaSYesYes-
9MidSMB12552FinTechYesNo-
EXPLAINABLE AI

SHAP Feature Attribution

Calculates how much each column contributes to the prediction using local cooperative game theory weights.

Awaiting local model calibration...

Click Train & Analyze in the sidebar to run WebGPU inference and generate SHAP importance.
Global Feature Importance (OfferDiscount) Pushes Higher Pushes Lower
CLV (High)+42.1%
ActivityScore (High)+28.5%
SupportTickets (High)-19.4%
Segment (SMB)-11.2%
MCP PIPELINE INTEGRATION

Orchestrate Autonomous Agents

Exposing sheet models to basic chat is just the beginning. The real power is connecting your model to autonomous agent loops (like background operations or decision-making pipelines) via MCP. Click a prompt below to see how an automated agent uses your uploaded models to run predictions on incoming profile data.

Claude Desktop • Connected to Carla HQ MCP

Active MCP Servers: Carla HQ (v1.0.4)

Exposing Tools: get_prediction, get_model_details, run_forecasting

T

Hello! I am connected to your Google Sheets models via Carla HQ. What predictions or analysis would you like me to run?

Select a query to send to Claude:
Quick Navigation