Hone: A Local-LLM Spreadsheet Tool Built on Decent
Hone is a web app that integrates local Large Language Models (LLMs) into spreadsheet workflows. It runs entirely in the browser—no servers, no data leaks—making it a practical example of the Decent architecture for local-LLM web apps.
Key Features
- Spreadsheet Import/Export – Supports Excel (XLS, XLSX), CSV, and TSV. Copy-paste from Excel, Google Sheets, Numbers, LibreOffice, and others.
- Prompt Templates – Define questions for each row; results populate back into the sheet.
- Local LLM Execution – Uses Llama 3.2 8b directly in the browser.
- Efficient Job Control – Start, pause, and resume LLM tasks as needed.
- Handles Large Data – Up to 256 columns and thousands of rows (device memory permitting).
- Fully open source - distributed under the permissive MIT license
Custom LLMs and Container Hosting
You can run Hone from our website , and that's the easiest way to use it. But you can also run your own instance using our Docker image. This deployment option allows for configuring the app to connect to LLMs served from outside the browser - for example, Ollama running on an on-prem server, or a corporate inference endpoint hosted from Azure or AWS. See the deployment guide for details
Why Hone?
Hone isn’t just a spreadsheet tool—it’s a real-world proof-of-concept for the Decent architecture. It demonstrates how local-LLM web apps can be private, performant, and fully self-contained.