I built a free local dashboard with Claude Code that shows what each prompt uses in tokens and API-equivalent dollars

I built CCUsageX, a free and open-source dashboard for Claude Code usage. It runs on your own computer and nothing is uploaded. What it does: it reads the transcript files Claude Code already writes to ~/.claude/projects and shows, for each prompt you type, the API requests it caused (subagents included), tokens, cache hits, response time and API-equivalent cost. It also has a live view of what Claude Code is doing right now, your 5-hour and weekly limits with a pace projection, per-model, per-session and per-project breakdowns, and CSV export. Why I made it: I use Max and kept running tight on limits, but I could not see which prompts were using them up. How Claude helped: I built it with Claude Code. I described what I wanted and checked the results, and Claude Code wrote nearly all of the code, including the log parser, the web UI, the live view and the tests. One thing I learned: Claude Code's logs write one line per content block, so counting every line as a request gives wrong totals. Lines from the same request have to be deduplicated. We checked the token counts against ccusage to catch this kind of mistake. Free to try: MIT license, no paid tier, no npm dependencies, needs Node 18+. git clone github.com/MHJoy99/CCUsageX.git cd CCUsageX node server.js Then open 127.0.0.1:3020 Demo with fictional data: mhjoy99.github.io/CCUsageX/demo Repo: github.com/MHJoy99/CCUsageX The costs are API list-price equivalents, so on a subscription they show what the usage is worth, not what you are billed. Feedback is welcome, especially if your numbers look wrong.