AI Credits usage

Hi Nancy, there isn’t a fixed credit cost per frame. Credits are consumed per agent interaction based on how much the agent has to read and generate, and frames are expensive because every prompt makes the Modeler agent write or rewrite a sizeable chunk of code, plus it loads the context of the app it’s working in. Iterative “tweak this, now change that” loops add up fast because each turn regenerates a lot.

A few things that helped me keep it down:

• Build frames in a small standalone app and share only the blocks you need into it, so the agent isn’t carrying your whole model as context each time.
• Write a detailed spec up front (layout, data blocks, interactions) and get as close as possible in one or two prompts, rather than many small iterations.
• Reuse existing frame code where you can. Nathan shared his Slide it frame with a skill.txt here on the community, and feeding the agent working code plus instructions is far cheaper than asking it to invent from scratch.
• Once the frame is close, make small edits to the code directly instead of asking the agent.

You can see consumption in the workspace AI usage page, and if you’re hitting your allocation it’s worth raising with your CSM since allocations differ by plan.