GRID

The interesting thing about GRID isn't that it reads and writes cells — plenty of libraries do that. It's that it collapses the agent loop. Without it, an LLM writes code, you execute it, you hope it didn't mangle the formula graph. With it, the model makes a function call and the engine recalculates the way Excel would. We'd reach for this the moment spreadsheet automation stops being a demo and becomes a feature you have to support. The four-product split is the practical part. You don't have to buy the whole stack. If you already have a UI and only need agent-accessible workbook math, take the engine plus agent tools and ignore the viewer. If you're building a customer-facing tool and want a spreadsheet on screen, the React Viewer and Editor save you from writing your own grid renderer, which is a genuinely miserable week of work. Where it earns its keep: financial models with interdependent formulas, anything where a wrong cell silently poisons the output, and agent products where the token bill is getting embarrassing. The Excel add-in route matters too — running domain skills on the engine instead of letting a model act on a live file is exactly the kind of sandboxing regulated industries want before they let an LLM near a model. Where it bites. This is a component, not a destination. There is no consumer spreadsheet to open on a Tuesday and start typing in. Teams without engineers will bounce off npm entirely. And if real-time multi-user co-editing is on your requirements list, that's a different product category. On price: the vendor page confirms a free evaluation license on npm install and commercial licensing when you ship. Beyond that, licensing is a conversation. Go in knowing your seat count and deployment shape before the call. Closest alternative