QAD — building a CAD as AI-first from the start

QAD is our experiment with a slightly different approach to CAD.

Instead of taking the existing way of working with CAD and adding AI to it later, we are building the tool as AI-first from the start.

For now we are at an early prototype stage. What we are mainly checking is the direction itself, and what an engineering tool should look like when it can be used not only by a person, but also by an agent.

A CAD available over MCP

One of QAD's basic assumptions is the use of MCP.

MCP itself may sound fairly technical, but from the user's point of view the idea is much simpler: you can design by talking to a chat and giving it the data it should work on.

QAD is a tool an agent can use over MCP.

So it is not only about talking to an AI about a project. The agent gets access to a tool in which it can do the work.

Hence our main assumption:

an agent that can use MCP can become an engineer and a CAD designer.

That is what we are now trying to check in practice.

Designing through conversation

This way of working is one of the most interesting parts of the whole idea for us.

Instead of being limited to working directly in the CAD, you can tell the agent what you want done and give it the data it needs.

The interaction can therefore happen through conversation.

That is also why we are building QAD as AI-first from the very beginning, with MCP as one of its basic assumptions.

We are currently testing QAD on injection moulds

One of the areas we are testing right now is injection moulds.

The screenshot attached to this post comes from one of those tests.

QAD is still an early prototype, so at this stage we are mainly building, checking our assumptions and watching how the approach works in practice.

What we want to find out

The most important question for us right now is not:

“How many CAD features can we build?”

We are more interested in:

“What should a CAD look like if we assume from the start that one of its users will be an agent?”

And that is the question we are now trying to answer.