I spent last week logging every Claude Code run I made in Framer, mostly because I had no idea what any of it was actually costing me.
The big tasks turned out not to be the problem. The vague ones were.
Copy fixes: approximately $0.50 – $5
Section rebuilds: $4–$16
Total: 40+ prompts over ~18 hours, roughly $139 at API-equivalent pricing
My real spend was nothing beyond the $20 Claude subscription I already pay for, which is its own kind of interesting. But the thing that stuck with me was how fast a lazy prompt compounds. I'd ask for something half-formed, get something half-right, correct it, correct it again. Three runs to do one thing.
Being specific the first time is cheaper than being fast.
Disclosure: I built Dexter, which is the only reason I have these numbers at all. It puts Claude Code beside the Framer canvas — select a layer, describe the change, and see the estimated cost when it lands.
If you're running external agents with Framer: do you track your usage, or do you only find out when you hit your limit?
I spent last week logging every Claude Code run I made in Framer, mostly because I had no idea what any of it was actually costing me.
The big tasks turned out not to be the problem. The vague ones were.
Copy fixes: approximately $0.50 – $5
Section rebuilds: $4–$16
Total: 40+ prompts over ~18 hours, roughly $139 at API-equivalent pricing
My real spend was nothing beyond the $20 Claude subscription I already pay for, which is its own kind of interesting. But the thing that stuck with me was how fast a lazy prompt compounds. I'd ask for something half-formed, get something half-right, correct it, correct it again. Three runs to do one thing.
Being specific the first time is cheaper than being fast.
Disclosure: I built Dexter, which is the only reason I have these numbers at all. It puts Claude Code beside the Framer canvas — select a layer, describe the change, and see the estimated cost when it lands.
If you're running external agents with Framer: do you track your usage, or do you only find out when you hit your limit?
