Accuracy issue

[PH010546]

First of all, thank you for this amazing tool!

I'm now using ccusage v18.0.10 with claude code v2.1.87, which has the /usage panel.

I found out that the claude code usage stats is much higher than ccusage gives.

Claude code usage shows that in the current seesion I used 22%, meanwhile ccusage shows I used only 6.9%.

Is there any config that I missed?

Thanks

[Astro-Han]

The gap is likely because ccusage derives usage from JSONL token counts, while CC's /usage display reads from its internal quota tracker. These measure different things: JSONL captures what was logged to disk (which can undercount for reasons discussed in #866 and #901), while the quota tracker reflects what Anthropic charges against your plan.

For numbers that match CC's /usage, the statusLine stdin JSON exposes rate_limits.used_percentage from the same internal source. That's how claude-pace stays in sync with CC's display.

[PH010546]

Thanks for answering! I tried statusline right away, but it shows 'No imput provided'

[Astro-Han]

The "No input provided" error means ccusage's statusline command isn't receiving JSON from Claude Code via stdin. This usually happens when the command is run directly in a terminal instead of through Claude Code's statusline hook.

Make sure your ~/.claude/settings.json has the statusLine config as documented:

{
  "statusLine": {
    "type": "command",
    "command": "bun x ccusage statusline",
    "padding": 0
  }
}

Then restart your Claude Code session. Claude Code pipes session data to the command on each assistant message, so it won't work outside that context.