Everyone Is Arguing About Claude Code Limits. Almost Nobody Knows Their Own Number.

How to check your real Claude Code and Codex usage in dollars, even on a Pro or Max subscription where the app won’t tell you.

If you follow any developer community right now, you’ve seen the arguments. Claude Code limits are too tight, or too generous. The $200 Max plan is a steal, or a ripoff. Someone’s company burned through its AI budget in a quarter. Someone else swears they get thousands of dollars of API-equivalent work out of a flat subscription every month.

What’s striking about these debates is how few people in them can cite their own usage. Not the plan they’re on. What they actually consume.

There’s a reason for that, and it isn’t laziness.

Subscription users can’t see their costs

If you pay for Claude through the API, you get a billing dashboard and the /cost command inside Claude Code. Clear numbers, per session.

If you’re on Pro or Max, like most heavy users, you get neither. The /usage command shows you a percentage of your rate limit, which tells you how close you are to being cut off but nothing about the value of what you've consumed. Your Anthropic console shows API billing, which is zero, because you're on a subscription. Token cost visibility for subscription users has been one of the most requested features in Claude Code's issue tracker for a long time now.

So the loudest pricing debate in the developer world is happening almost entirely on vibes. People are deciding whether to upgrade, downgrade, or switch providers based on how their usage feels.

The data exists. It’s on your laptop.

Here’s what gets missed in these discussions: Claude Code logs every session locally, in JSONL files under ~/.claude/projects/. Each entry records the model, the input and output tokens, cache reads and writes, the timestamp, and the working directory. Codex CLI does the same under ~/.codex/sessions/, and so do Cursor, Windsurf, Cline, Aider, and most other AI coding tools.

Multiply those token counts by the published per-million-token API prices and you get the number everyone in these threads is missing: what your usage would cost at API rates, in real dollars, per day, per project, per model.

You could script this yourself in an afternoon. Or you can use AI Usage Tracker, a free, open source macOS app I built that does it automatically.

What you’ll see

The app scans the local logs of ten-plus tools, covering both Claude and Codex, and renders one dashboard. No account, no cloud, no telemetry. Your logs contain fragments of your conversations and your file paths, so everything runs on your machine and nothing is uploaded anywhere.

Once it’s open, the pricing debate gets a lot more concrete for you personally:

• Your all-time and monthly API-equivalent spend, next to what your subscription costs. This settles the “is Max worth it” question in about ten seconds. For some people the answer is an easy yes. For lighter users it’s genuinely not, and that’s worth knowing before the next renewal.
• A per-model breakdown, so you can see how much of your spend is an Opus-class model doing work a cheaper model could handle.
• Cost grouped by project, using the working directory in each log entry. Most people find one repo quietly consuming several times more than everything else.
• A daily chart and an hour-by-day heatmap, which together show exactly when and where the burn happens.

There’s also a session log with full token detail and a one-click claude --resume or codex resume command for any past session.

Your number changes the argument

I’m not going to tell you whether the current limits are fair or whether the subscription is a good deal. That genuinely depends on your usage, which is the whole point.

A developer consuming $900 of API-equivalent tokens on a $200 plan is getting a bargain and should probably stop worrying. A developer consuming $60 on the same plan is paying for peace of mind they may not need. Both of them are in the same comment threads right now, arguing past each other, because neither has looked.

The tool is free at https://usages.pro/. Build it from source, or grab the signed macOS build for a one-time $9 if you’d rather skip the build step. Either way, the next time the limits discussion flares up, you’ll be one of the few people in it who actually knows their number.