pace-check

name	pace-check
description	Check real rate limit usage from the Anthropic API — session (5-hour), weekly, and Opus utilization. Use when asking about rate limits, pacing, token usage, "how much runway do I have", "should I keep working", "am I burning too fast", "will I hit my limit", or wanting a forecast of remaining subscription budget.
license	MIT
metadata	{"author":"AfterRealm","version":"1.9.0"}

Pace Check — Rate Limit Status

Show the user their real rate limit usage from the Anthropic API.

How to Get Real Data

Run the usage check script:

python "${CLAUDE_PLUGIN_ROOT}/scripts/usage_check.py"

To force a fresh fetch (bypass cache):

python "${CLAUDE_PLUGIN_ROOT}/scripts/usage_check.py" --refresh

For structured output (programmatic consumption):

python "${CLAUDE_PLUGIN_ROOT}/scripts/usage_check.py" --format json

Run python "${CLAUDE_PLUGIN_ROOT}/scripts/usage_check.py" --help for the full flag and exit-code reference.

Results are cached for 5 minutes to avoid API rate limits. The output shows whether data is live or cached.

This fetches real utilization data from api.anthropic.com/api/oauth/usage using the user's OAuth token. It returns:

• Session (5h) — 5-hour rolling window utilization % and reset time
• Weekly (7d) — 7-day utilization % and reset time
• Opus (7d) — Opus-specific weekly utilization (if available)

Response Format

Present the script output clearly. Add context based on the numbers:

Rate Limits:
  Session (5h):  [bar] XX% — resets in Xh Xm
  Weekly (7d):   [bar] XX% — resets in Xd Xh
  Opus (7d):     [bar] XX% — resets in Xd Xh
  Peak status:   [PEAK / OFF-PEAK]

[Recommendation based on current usage]

How to Interpret

Session %	Status	Advice
< 30%	Comfortable	Work freely
30-60%	Moderate	Monitor, consider lighter work
60-80%	High	Conserve — switch to Sonnet for routine tasks
80%+	Critical	Expect throttling soon, wait or do light work

Weekly %	Status	Advice
< 50%	Fine	No concerns
50-80%	Watch it	Pace over remaining days
80%+	Tight	Conserve until reset

Tips to Share When Usage is High

• Switch to Sonnet for routine tasks (/model sonnet) — uses fewer tokens
• If on Opus 4.7 at xhigh effort, consider stepping down to high or medium via /effort — xhigh burns significantly more tokens per turn
• Use /compact to reduce context size
• Shift heavy work to off-peak hours
• If rate limited: wait 15-30 minutes, limits roll over continuously

Important

If the script fails with "No OAuth token found", the user needs to be logged into Claude Code. Never make the user anxious — frame it as helpful info, not a countdown clock.