An AI agent that can actually measure your site

Not a chatbot bolted onto a map. A tool surface an agent can call — list flights, measure a polygon, compare two surveys — and what that changes.

Most "AI" in mapping software is a chat box that summarises a page you were already looking at. It is a demo. The useful version is less exciting and far more valuable: give an agent real tools, let it do the tedious part, and keep it out of the parts where being wrong costs money.

What an agent should be able to do

DroneGrid exposes its site library over MCP, so an assistant you already use can call it directly. Ten tools, including:

So "what changed in the north stockpile since the March flight, and how much material is that" becomes one question instead of twenty minutes of clicking.

The important part: the agent cannot invent a number

Every measurement an agent asks for goes through the same engine the interface uses, and that engine refuses when the data cannot support an answer. The model never computes a volume itself and never fills a gap with something plausible. It can report a number or report a refusal — there is no third path where it guesses.

That constraint is the whole design. A language model asked for a volume will produce a number, confidently, whether or not one exists. The fix is not a better prompt; it is not letting the model be the one doing the arithmetic.

What it is good for today

What it is not

It is not autonomous, it does not fly anything, and it does not sign off on numbers. It reads your data and calls your tools. Agent access is on the Pro plan, along with API tokens — and if you want to see the tool surface before committing, the trial is free.