HN Showcase
Back to browse
GitHub Repository

Tracking Claude Code Usage Limits

87 starsPython

I tried to reverse engineer Claude Code usage limits

by aray07·Mar 26, 2026·5 points·0 comments
Visit ProjectView on HN

AI Analysis

●●SolidSolve My ProblemDark Horse

Exposes hidden rate-limit headers Claude Code never shows you.

Strengths
  • •Captures anthropic-ratelimit-unified headers with async logging that stays ahead of disk writes.
  • •Normalizes SSE streams, token counts, and cache reads into queryable JSONL.
  • •Produces rough budget bands ($78–$282 per 5h) from actual traffic patterns.
Weaknesses
  • •Alpha stage — estimators produce ranges, not precise limits you can depend on.
  • •Only works for Claude Code users; niche audience limits broader appeal.
Target Audience

Claude Code subscribers hitting opaque usage limits

Similar To

Warp · Cursor · Continue

Post Description

I hit my Claude Code usage limit yesterday in the middle of regular work. Not doing anything unusual - just a couple of coding sessions. It felt like it came out of nowhere.

Then today Anthropic announced they're tightening 5-hour limits during peak hours. It feels like a rug pull is coming and we're all just supposed to accept "you've used too much" with no explanation of what "too much" means.

Usage limits have always been opaque. A vague percentage bar. No dollar figure. No formula. No way to plan your work around it.

So I built a proxy that captures the rate-limit headers Anthropic sends back on every response — headers that exist but Claude Code never shows you.

Here's what my data looks like for the last 2 days:

My 5h budget is worth ~$78–$282 Estimated 5h Budget (11 sessions observed) Range: $34 – $399 Median: $164 p25-p75: $78 – $282

Estimated 7d Budget Estimate: ~$1,949

I'm paying $200/month for Max 20x. My estimated weekly budget is ~$1,949 in API-equivalent usage.

The quota is price-weighted, not token-based 95.1% of my input tokens were cache reads (427M tokens). They barely moved the quota.

The 5.5M fresh input and 1.8M output are what actually burned through the budget (as expected).

If you're on Pro or a different Max tier, your data would help a lot.

It’s open source so more people can run it and we can figure out the real numbers together.

https://github.com/abhishekray07/claude-meter

Similar Projects