{"slug": "show-hn-my-claude-quota-ran-out-in-10-minutes-so-i-made-a-tool-to-find-out-why", "title": "Show HN: My Claude quota ran out in 10 minutes, so I made a tool to find out why", "summary": "Developer kelviq released Tare, an open-source tool that lets Claude Code users query its own logs in plain English to diagnose why their usage quota is draining, after the creator's own Claude quota ran out in 10 minutes. The tool, installed via a single npx command, answers questions like 'Why did I hit my usage limit yesterday?' and attributes token consumption to specific projects, models, or background sessions, with commands like '/tare' for a full diagnosis and '/tare report' for an HTML report.", "body_md": "**Ask Claude Code where your usage went.**\n\nYou hit a usage limit and don't know why. Your quota drains faster than it\nused to. You suspect something is eating tokens in the background. The\nrecords that answer all of this are already on your computer — Claude Code\nkeeps a log of every request it makes. `tare`\n\nteaches Claude Code to read\nits own logs, so you can just ask.\n\nNo dashboards, no commands to learn. Install once, then ask in plain English.\n\n*Tare: the weight of the container, subtracted to find what's inside. Most\nof what a session costs is the container — context re-sent again and again —\nand that's exactly what these tools subtract.*\n\nOne command, in any terminal:\n\n```\nnpx skills add kelviq/tare -g -y --copy --agent claude-code\n```\n\nThat's it. Nothing else to set up — no accounts, no packages, no\nconfiguration. Start a **new** Claude Code session and it's live — type\n`/`\n\nand check that `tare`\n\nappears. If it doesn't, see the\n[troubleshooting note](/kelviq/tare/blob/main/INSTALL.md#troubleshooting) — on machines where\nClaude Code has never installed a skill before, the installer can miss\nit. (Other ways to install — by hand, for a whole team, or as a Claude\nCode plugin — are in [INSTALL.md](/kelviq/tare/blob/main/INSTALL.md).)\n\nOpen Claude Code and ask your question the way you'd ask a person. These all work — the words don't have to match, complaining about your limits is enough:\n\n**When you hit a limit**\n\nWhy did I hit my usage limit yesterday?\n\nI got locked out ten minutes into my evening session — how is that possible?\n\nDid I hit the 5-hour limit or the weekly cap?\n\nWhy am I burning through my quota so much faster this week?\n\n**Where your tokens go**\n\nWhere did my tokens actually go this week?\n\nWhich of my projects is eating my quota?\n\nWhich model is costing me the most?\n\nWhat's the most expensive file Claude keeps re-reading?\n\nHow much did that giant session yesterday actually cost me?\n\nAre my MCP servers adding a lot to my context?\n\nHow much overhead do subagents and skills add?\n\n**Before you start something big**\n\nHow full is my 5-hour window right now?\n\nIs it safe to start a big refactor now, or should I wait for my window to clear?\n\n**Checking for things you forgot**\n\nIs something running Claude Code in the background?\n\nWas Claude Code active while I was asleep?\n\nI set up an automation last month — what is it costing me?\n\n**Tuning your setup**\n\nI started using /clear between tasks — did it actually help? Compare this week to last.\n\nDid the latest Claude Code update change my usage?\n\nWhat would my usage cost if I were paying for the API directly?\n\nWhat one change would save me the most?\n\n**Reports and sharing**\n\nMake me a usage report I can open in my browser.\n\nGive me a summary I can post publicly — with nothing private in it.\n\nExport my usage to a spreadsheet.\n\nPrefer typing commands? `/tare`\n\nruns the full diagnosis, and takes variants\nfor the common asks — asking in plain words always works too:\n\n`/tare` |\nfull diagnosis — where tokens went and why |\n`/tare usage` |\nat-a-glance panel — like `/usage` , with attribution |\n`/tare window` |\nhow full is the 5-hour window — safe to start? |\n`/tare report [days]` |\nbuild the HTML report and open it |\n`/tare tools [days]` |\nwhat is filling my context |\n`/tare week` |\ncompare this week with last |\n`/tare share [days]` |\nredacted summary safe to post publicly |\n`/tare why did I hit the limit yesterday` |\nany question works as the argument |\n\nNot a wall of numbers — a cause. The answer to \"why did I hit my limit yesterday?\" looks like this:\n\n99% of yesterday's usage came from a tool you're running, not from you.Something spawned 1,553 short Claude Code sessions in your website project — 9,022 requests, up to 51 sessions running at once. Your own hands-on work that day was 93 requests. Each fresh session rebuilds its context from scratch, which is the most expensive way to spend tokens...\n\n...followed by the evidence, what to check, and what to change. And when everything is actually fine, it says that: usage proportionate, no anomaly, here's what's normal for you.\n\nReal output lives in [examples/](/kelviq/tare/blob/main/examples) — a shareable summary produced\nby `/tare share`\n\n, and the HTML report.\n\n**Correct totals.** Claude Code's log format repeats each API response several times over; naive counting inflates totals — by 86% on the data this was built against. tare deduplicates properly.**The real cost of context.** A file read early in a long session gets re-sent with every later message. tare charges tools for what they*caused*, not just what they returned — which is how one big file read early can quietly dominate a week.**The rolling window.** Limits don't reset when you walk away; work from four hours ago still counts. tare can tell you how full your window was at the exact moment you were locked out.**The shape of automation.** Hundreds of short parallel sessions is a script, not a person. tare recognises the signature and says so.\n\nEverything runs on your machine and nothing leaves it. The scripts make no network connections at all. When you ask for a shareable summary, it contains totals, dates and tool names only — no prompts, no file paths or contents, no commands, no session or account identifiers — so you can post it publicly or send it to a colleague and ask \"what am I missing?\"\n\nDon't take that on faith: [SECURITY.md](/kelviq/tare/blob/main/SECURITY.md) states exactly what\neach file reads, writes and sends, and shows how to verify every claim\nyourself with one grep.\n\n- Claude Code on macOS or Linux\n- Python 3.9+ — already present on every Mac; no packages to install\n- Currently reads Claude Code's logs only, not other coding agents'\n\n## tare-screen-recording.mp4\n\nEverything the skill does, you can also do by hand: three dependency-free\nPython scripts with recipes for scripting, cron, CSV export, live\nper-request telemetry and more — see [CLI.md](/kelviq/tare/blob/main/CLI.md).\n\nIssues and PRs welcome. Useful directions: a live TUI, Windows paths, aggregating anonymised summaries across users to spot patterns no single person can see, and better token estimation for tool results.\n\nMIT licensed.", "url": "https://wpnews.pro/news/show-hn-my-claude-quota-ran-out-in-10-minutes-so-i-made-a-tool-to-find-out-why", "canonical_source": "https://github.com/kelviq/tare", "published_at": "2026-08-27 16:37:44+00:00", "updated_at": "2026-08-27 16:49:35.015127+00:00", "lang": "en", "topics": ["developer-tools", "ai-tools", "artificial-intelligence"], "entities": ["kelviq", "Tare", "Claude Code", "Anthropic"], "alternates": {"html": "https://wpnews.pro/news/show-hn-my-claude-quota-ran-out-in-10-minutes-so-i-made-a-tool-to-find-out-why", "markdown": "https://wpnews.pro/news/show-hn-my-claude-quota-ran-out-in-10-minutes-so-i-made-a-tool-to-find-out-why.md", "text": "https://wpnews.pro/news/show-hn-my-claude-quota-ran-out-in-10-minutes-so-i-made-a-tool-to-find-out-why.txt", "jsonld": "https://wpnews.pro/news/show-hn-my-claude-quota-ran-out-in-10-minutes-so-i-made-a-tool-to-find-out-why.jsonld"}}