skill

claude-usage-pacer

Claude Code skill: meter your 5-hour and weekly Claude plan usage and pace work to land near the cap at each reset

One-Lyf/claude-usage-pacercommunitylistedunreviewed

Set up

Skills are installed by placing their folder where your agent loads skills.

Download ZIP from GitHub Open the repository

The download comes from github.com/One-Lyf/claude-usage-pacer itself, not from our server, so what you get is exactly what its author published.

Install instructions come from the README

Skills do not use an MCP config. Follow the steps in the README below.

README

synced Oct 25, 2026

claude-usage-pacer

A Claude Code skill that meters your Claude plan usage (5-hour and weekly limits) and paces work against them, so long or parallel sessions use the plan fully without hitting the wall.

Claude checks live usage itself, compares it to a pace line, and decides whether to add a subagent, hold, ease off or stop. No more finding out at 100%, and no more resets with half the window unused.

5h  used  15%  target  63%  BEHIND     resets in 1h28m (Fri 6:00 PM)  room 50.7%/h
7d  used  41%  target  39%  ON PACE    resets in 3d 22h (Tue 3:00 PM)  room 0.5%/h
advice: SPEED UP: capacity is going unused. Add one parallel task or do more yourself.

Install

git clone https://github.com/One-Lyf/claude-usage-pacer
mkdir -p ~/.claude/skills
cp -R claude-usage-pacer/skills/usage-pacer ~/.claude/skills/

Start a new Claude Code session. Claude loads the skill when pacing matters, or you can ask: "check my usage" or "pace this work to my limits".

Run the meter yourself any time:

python3 ~/.claude/skills/usage-pacer/usage.py

Requirements

Claude Code logged in with a Claude subscription (Pro or Max). API-key logins have no plan limits to meter.
Python 3.7+ (stdlib only).
macOS (Keychain) or Linux/Windows (~/.claude/.credentials.json).

Configuration

Variable	Default	Meaning
`PACER_CAP_5H`	90	Ceiling for the 5-hour window, percent
`PACER_CAP_7D`	90	Ceiling for the weekly window, percent
`PACER_TOLERANCE`	5	Points either side of the pace line counted as on pace

How it works

The script reads your Claude Code login token locally and calls Anthropic's account usage endpoint for your 5-hour and weekly utilization. For each window it computes target = cap x elapsed / window length and reports whether you are behind, on or ahead of that line. The weekly window is treated as the binding limit.

Privacy and caveats

The token is read on your machine, sent only to api.anthropic.com, and never printed or logged. Output is percentages and reset times only. Read usage.py yourself; it is under 200 lines.
The usage endpoint is undocumented. It can change or break without notice. If it does, the skill tells Claude to fall back to asking you to run /usage.
Not affiliated with or endorsed by Anthropic.

License

MIT