Theo (t3.gg) was the first person I heard say the Max plan was actually worth it. I took his word for it, started using Claude Code more, and kept climbing the tiers — Pro → Max 5x → Max 20x — because each one felt like it was paying for itself.
But felt like is doing a lot of work in that sentence. Was the value real, or was I rationalizing the spend because I wanted it to be?
Recently, someone on a Reddit thread put words to that doubt: I was probably comparing API list prices to retail and lying to myself about savings. So I built a dashboard to actually check.
Three and a half months: 6.4B tokens · 29,269 messages · 57 projects · $292 paid on subscription · ~$4,426 equivalent at API list prices · 15.2× cheaper on that naive comparison.
The Reddit argument isn't entirely wrong — list prices aren't necessarily what you'd actually pay if you batched, used cheaper models, or negotiated. But I'm not going to pretend I'd have run my biggest project for 24 hours straight at API rates out of pocket. The honest version: the subscription made experimentation possible at all. The savings are real because the alternative wasn't "API at list" — it was "don't build it."
The dashboard tracks tokens, API-equivalent cost, and what you've actually paid — fully local, nothing leaves your machine. Optional Gmail connect for real invoices, or skip that and enter your tier manually. One command to run; data loads in about ten seconds and stays fast after that. I aimed for the middle ground — not so technical it scares people off, not so locked-down you can't peek under the hood. Built in maybe two hours, vibecoded, bad name — it answers the question for me anyway.
github.com/ajmalaksar25/claude-usage-dashboard
README → full write-up — setup, scope, and anything I'd spell out in a long post.
If your numbers come out very different from mine, tell me. I'd rather know than keep guessing.
BuildingInPublic · ClaudeCode · DeveloperTools · AIEngineering