Why We Built a Daily Credit Pool Dashboard for Claude Code — And the $100/Mo Cap That Changed Our Budget

TL;DR: Anthropic’s credit pool ($100/mo for ~100 Sonnet-equivalent credits/day) turned our unpredictable $287/mo AI spend into a fixed $100 line item — but only if we stay in pool. We cron a daily 5pm check and Slack-alert at 85% consumption. Full pricing analysis →

The Context

Two-dev team, ~120 hrs/mo AI-assisted coding across Next.js, FastAPI, and Terraform. Pre-June stack: Cursor Pro ($20) + Anthropic API direct (Opus-heavy, $150–300/mo). May 2026 actual bill: $287. Credit pool announced May 20, hard cutoff June 15: $100/mo covers ~100 credits/day (Sonnet), Opus at API rates after pool exhausted.

What We Tested

Configuration	Monthly Cost (Est.)	Verdict	Why
Cursor Pro + Anthropic API (Pre-June)	$287 (actual May)	❌ Baseline	Unpredictable; Opus spikes on auth/infra
Claude Code Credit Pool ($100/mo)	$100 (projected)	✅	Fixed cap; Sonnet covers 85% of our tasks
Credit Pool + explicit Opus flag	$100 + ~$40 Opus	⚠️ Acceptable	Only for auth/infra/payments; tracked via --model opus + reason flag
Cursor Pro only	$20	❌	Composer context loss + no autonomous loops
Codex (ChatGPT Plus)	$20	❌	Different mode; no terminal autonomy

The Pivot Point

May 28 dry-run on pool math: 22 working days × 8.5 hrs/day. At ~$0.60/hr effective (Sonnet), that’s ~$112 — slightly over pool. But 30% of hours were Opus (auth/infra). Realization: Pool isn’t per-token; it’s a daily credit budget (~100 credits = ~$5 Sonnet-equivalent). We must stay in pool daily. Built 5pm cron alert at 85 credits (see dashboard below).

What We Use Now

Daily Spend Dashboard (.toolcrucible/claude-spend-dashboard.sh, runs via cron 5pm daily):

#!/bin/bash
# claude-code usage --json | jq '.daily_credits_used, .daily_usd_estimate'
USAGE=$(claude-code usage --json --since yesterday)
CREDITS=$(echo "$USAGE" | jq '.credits_used')
USD_EST=$(echo "$USAGE" | jq '.usd_estimate')
THRESHOLD=85  # 85% of 100 credit pool
if [ "$CREDITS" -gt "$THRESHOLD" ]; then
  slack-post "#dev-alerts" "⚠️ Claude Code credits: $CREDITS/100 (~$${USD_EST}). Opus flag review needed."
fi
echo "[$(date)] Credits: $CREDITS/100 | Est: $${USD_EST}" >> ~/.toolcrucible/claude-spend.log

Team Protocol:

• Default model: sonnet (in pool)
• cc --model opus requires --reason "auth|infra|payments" flag (logged)
• Weekly review: claude-code usage --since 7d --by-model → adjust Opus allocation
• If pool exhausted before day-end: fallback to Cursor for remainder (happened 2× in first 3 weeks)

When You’d Choose Differently

• Light users (<10 hrs/week): Credit pool overkill; pay-per-token API or Cursor Pro cheaper.
• Opus-heavy workflows (>50% Opus): Pool economics break; direct API or competitors (OpenAI o1-pro, Gemini 2.5 Pro) may win.
• Enterprise/teams >10: Anthropic enterprise plans (unannounced as of Jun 2026) needed; pool is per-seat.

Tool Crucible Rating

Dimension	Rating (1–5)	Notes
Overall	4.5	Best pricing innovation 2026; forces discipline
Ease of Use	4	claude-code usage CLI solid; dashboard DIY
Value	5	60%+ savings vs pre-June stack for heavy users
Support	3	Discord only; billing edge cases slow

---

This is part of our AI Coding Tool Evaluation series. See full analysis: Claude Code June 2026 Credit Pool: Pricing Deep Dive & Dashboard