Configure your usage
Adjust the sliders to match how you use Claude. Estimates assume prompt caching is active (which Claude Code uses automatically).
3
1 (casual)
10 (heavy)
20 (all-day)
100K tokens
10K (quick chat)
100K (coding session)
500K (long project)
70%
0% (no cache)
70% (typical Claude Code)
95% (large codebase)
Claude Code caches CLAUDE.md, MCP schemas, and system prompts automatically. Higher cache rate = cheaper API costs.
API cost
$0
per month at API rates
Max plan
$100
flat monthly rate
Unlimited sessions · No per-token billing
Savings
$0
per month with Max
Your usage
Adjust the sliders to see your estimate.
API cost vs Max plan
0%
$0 API cost
$100 (Max)
Break-even
—
Cost per session
—
Sessions/month
—
Tokens/month
—
Track it live in your terminal
Terminal setup guide →Add this snippet to your Claude Code statusline command to see the estimated API cost update in real time as you work — so you know exactly what you're saving each session.
# In your statusline command (receives JSON via stdin):
input=$(cat)
model_id=$(echo "$input" | jq -r '.model.id // empty')
total_input=$(echo "$input" | jq -r '.context_window.total_input_tokens // 0')
total_output=$(echo "$input" | jq -r '.context_window.total_output_tokens // 0')
cache_write=$(echo "$input" | jq -r '.context_window.current_usage.cache_creation_input_tokens // 0')
cache_read=$(echo "$input" | jq -r '.context_window.current_usage.cache_read_input_tokens // 0')
# Pricing lookup (per million tokens: input / output / cache-write / cache-read)
case "$model_id" in
*fable-5*) ip=10; op=50; cwp="12.50"; crp="1.0" ;;
*opus-5*|*opus-4-8*|*opus-4-7*|*opus-4-6*) ip=5; op=25; cwp="6.25"; crp="0.5" ;;
*sonnet-5*|*sonnet-4-6*) ip=3; op=15; cwp="3.75"; crp="0.3" ;;
*haiku-4-5*) ip=1; op=5; cwp="1.25"; crp="0.1" ;;
*) ip=3; op=15; cwp="3.75"; crp="0.3" ;;
esac
cost=$(awk "BEGIN {printf \"%.2f\", ($total_input*$ip + $total_output*$op + \
$cache_write*$cwp + $cache_read*$crp)/1000000}")
echo "~\$$cost"
Paste this into your ~/.claude/statusline-command.sh. Claude Code pipes JSON with cumulative token counts to the statusline command on every render.