CARLA HQ
Google SheetsTabICLv2100% Local AI

Predict churn, salaries, and category labels privately.

Train high-performance tabular foundation models locally in Google Sheets with Carla. Integrate them securely into your LLM workflows via Carla HQ using the Model Context Protocol (MCP).

PRODUCT OVERVIEW

Run Spreadsheet AI Locally. Query Privately. Save 99% on Token Costs

You shouldn't need a data science department to figure out what your spreadsheets are trying to tell you. And you certainly shouldn't have to ship your sensitive corporate data to a third-party cloud just to run a predictive model.

Carla operates entirely as a local sidebar extension. It analyzes your tabular columns, learns structural patterns, and generates highly accurate predictions for missing values—completely offline and local to your browser.

ONBOARDING

Zero-Friction Startup

01

Install Extension

Download Carla from the Chrome Web Store to add the side panel to Google Sheets.

02

Train Your Model

Highlight your features, select your target column, and train a model locally using WebGPU.

03

Local Inference

Run classification, regression, time series prediction and synthetic data generation right in your browser.

04

Upload to Carla HQ

Opt-in to securely upload your trained model weights to Carla HQ hosted servers, transitioning inference tasks to secure cloud container hosting.

05

Hook up Agent Loops

Connect autonomous agent loops via OAuth to the hosted Carla HQ MCP Server, enabling them to query and use your spreadsheet models in background workflows.

CORE INFRASTRUCTURE

Built for Speed and Security

Live Model Registry

Democratize ML across your team. Bypasses the complex engineering pipeline (Docker, Kubernetes, API Gateways). Analysts train locally and publish model parameters instantly as live API tools.

Privacy-First KV Cache

Uploads only compiled model weights (the KV cache) to Carla HQ, leaving actual table cell rows completely offline. Easily satisfies security teams worried about sensitive corporate data leaks.

Decision Gates for Agent Loops

Avoid the LLM-in-the-loop bottleneck. Hook up model predictions as fast decision gates inside autonomous workflows (e.g. auto-triggering discount codes in CRM pipelines) instead of basic text chat querying.

Direct REST API Gateway

Deploy your spreadsheet models as high-throughput REST APIs. Run fast, deterministic batch calculations directly in automated CRM or data pipelines without routing payloads through slow LLM loops.

Turn Your Static Spreadsheets into Live AI Infrastructure

Give your LLM models access to classification, regression, and forecasting without sending heavy database files.

Requires Chrome • Powered by TabICLv2 • No credit card required

Quick Navigation