Claude Code for your team: Max plans or API billing?

Every team adopting agentic coding argues about this. Most back of the envelope math ignores prompt caching and overstates API cost several times over. This calculator does the cache-aware version. Every assumption is visible and editable. No signup, no tracking.

Your team

Developers using Claude Code Usage intensity (fills the token assumptions below, then edit freely)

Lightunder 1 hr/day Daily driver2 to 4 hr/day Heavy6+ hr, parallel agents

Model mix: 25% Opus 4.8, rest Sonnet 5 Working days per month

Assumptions (the part most calculators hide)

Input MTok / dev / day

Output MTok / dev / day

Cache hit rate on input tokens: 85%

Agentic sessions re-read the whole conversation every turn, so most input tokens are cache reads at 0.1x the base price. Fresh tokens are modeled at the 1.25x cache write premium (5 minute TTL). Token volumes vary wildly by workflow; these defaults are rough medians from public usage reports, not gospel. Edit them.

Your team

Monthly cost per developer