I think it depends a lot on what model you use for the assistant. For example something like GPT-5.4 nano or mini are quite cheap, whereas if you are using the frontier models then yeah it can get pretty pricy. Usually it’s those frontier models that can really do a lot of the build and design work, whereas the cheaper models can be used for the simpler things.
@dan did a video here which should help BIG NEWS: Elements 3.0 is out of beta! 🎉 - #44 by dan
I’d have both, but probably use the MCP server option more often.