Is Archicad at a disadvantage for AI-assisted scripting compared to Revit and Rhino?

[Steffen]

I’m wondering whether Archicad is at a disadvantage for AI-assisted scripting compared to Revit. My impression is that the Revit API (and Dynamo) has far more public documentation and examples, which makes AI coding assistants much more reliable there. GDL is also a fairly niche language, and the Archicad Python API seems to cover fewer element types and operations so far. Is this something others have run into, or is the Archicad Python API maturing quickly enough that this gap is closing? And are there any workflows people use to get AI tools working well with GDL or the Python API?

Operating system used: Windows

[Mathias Jonathan]

I think that’s where we differ: I don’t see AI as hiding a shaky foundation, because the foundation isn’t what we’re building on with AI.

The C++ API already does most of the heavy lifting when it comes to connecting to Archicad’s core. What we do in Tapir, for example, is mostly framing inputs and outputs properly. Each command is a fairly thin layer on top of functions Graphisoft provides and maintains, so the room for serious errors is limited. Compilation errors are caught immediately, and AI is very good at reading and fixing them. Runtime issues show up quickly when testing, since each command does one specific thing. And as I mentioned, the code is reviewed by a maintainer who knows the API well. So yes, someone still verifies what AI produces. Verification just focuses more on behaviour and testing than on reading every line.

Graphisoft keeps full control over Archicad’s core and only exposes a fairly high-level access to it. So I don’t really see where the instability would come from. The foundation remains Graphisoft’s responsibility, and official and third-party add-ons have relied on this same API for years. What has changed is simply that it has become accessible to many more people.

I’ll ask again, because I’m genuinely curious: have you tried developing something yourself? What problems did you run into? Or is this more of a theoretical concern?

On GDL parameters and properties, I agree a direct read/write access from GDL would be a real improvement. But beyond technical debt, there are some fairly complex underlying questions: how properties stay in sync across projects, modules and hotlinks, across Archicad versions, and so on. GDL’s stability is also a strength. Objects created in very early versions of Archicad still work today and will likely keep working in the future. I don’t know many systems that can say the same. In an ideal world everything would be fully synchronized and consistent, but Archicad stays quite flexible on many levels, and that has value too.

To come back to your first message: the idea that the C++ API is too cumbersome to use with AI and that GDL is a limiting legacy system just doesn’t match my day-to-day experience.

[ooki]

You can automatically generate GDL objects using Claudecode.
By providing Claudecode with the GDL manual and some basic technical guidance, and having it reference the webpage for the temporary construction equipment you wish to create, the task can be accomplished with just two lines of script.
Although the video is fast-forwarded in parts, the entire automated generation process takes approximately 18 minutes.
The environment used is Archicad 29 with Tapir.

[Sefa Yakut]

Hi,

For a while now, I’ve been experimenting with Archicad’s AI capabilities using Claude and ChatGPT 6 Astra.

I can say that almost anything is possible. Archicad doesn’t just have a Python API—it also has a C++ API.

Since I can’t share a video of the dimensioning add-on I created using just a few natural sentences, I’m sending it as a GIF. In accordance with the dimensioning order used in my country (Turkiye), I can dimension all floors with a single click.

Archicad’s standard automatic dimensioning tool doesn’t provide dimensions between axes, but I’ve developed and am using an add-on leveraging Claude’s capabilities. It’s quite enjoyable.

Br,

Sefa