Show HN: My Claude quota ran out in 10 minutes, so I made a tool to find out why

Claude Code users often hit usage limits without understanding why—quotas drain faster than expected, background processes may be consuming tokens, and the detailed request logs on their machines remain unreadable without digging. Tare is a tool that teaches Claude Code to read its own logs, letting users ask questions in natural language instead of learning dashboard commands or SQL.

Installation is a single command: `npx skills add kelviq/tare -g -y --copy --agent claude-code`, with no accounts or configuration needed. After installing, users open a new Claude Code session and simply ask questions like 'Why did I hit my usage limit yesterday?' or 'Which project is eating my quota?' The tool also offers a `/tare` command with variants for common asks such as full diagnosis, at-a-glance usage panels, and specific reports.

Tare can answer questions about 5-hour limits versus weekly caps, token breakdowns by project/model/file, overhead from MCP servers, subagents, and skills, background Claude Code activity, the cost of automations, and the impact of setup changes like using `/clear`. It can also generate browser-viewable reports, sanitized public summaries, and spreadsheet exports.

The tool's name comes from the concept of tare weight: subtracting the container (re-sent context) to reveal what's actually inside. Most of a session's cost is the container—context re-sent repeatedly—and tare's purpose is to subtract that overhead to expose true token usage.