August 2026 — Added Claude Code
Claude Code is now the sixth platform on the site, with its own pricing guide, its own receipts on every cost page, and its own column in the tables above and below.
It bills differently from everything else here. There are no credits and no published allowance of tokens — the chunks of text the AI reads and writes. Claude Code comes with a Claude subscription — Pro at $20/mo ($17/mo billed annually), Max 5× at $100/mo, Max 20× at $200/mo — and the limits are rolling five-hour windows plus weekly caps, published only as multiples of the Pro plan. Run past a plan and you pay Anthropic's direct pay-as-you-go prices instead (Opus 5: $5 per million tokens read, $25 per million written; Sonnet 5: $3/$15, as of Aug 2026).
Because Anthropic publishes no allowance, the reference plan below carries an assumption of ours rather than a number of theirs: we model Pro as covering roughly $40/mo of build work — priced as if you were paying as you go — before the weekly caps bite, with Max at 5× and 20× that. It is an estimate, it is labelled as one on the methodology page and in the pricing guide, and it is the first thing we will correct if better information turns up.
Where it lands: on every default build we price, Claude Code comes fourth on cost, behind Bolt.new, Lovable and Cursor + Claude, and ahead of Replit and v0. No scissors icon changed hands. That is arithmetic, not an endorsement, and it cuts both ways: the gotcha still says you have to be able to read what it writes.
Correction, same day: this entry first placed Claude Code third, ahead of Cursor + Claude. That ranking came from a bug — we were adding Cursor Pro's $20/mo on top of the usage estimate when the $20 already buys that usage. Fixing the double-count moves Cursor + Claude ahead of Claude Code on every default build. The cheapest pick is unaffected.
The snapshot table below is generated live from the same data, so it already shows the six-platform line-up rather than the five we launched with.