$ git diff main -- thread/claude-code-review-launch-mar-2026.md
commit a1b2c3d - author: Oliver Wakefield-Smith - verified 21 Jun 2026
Anthropic launched Claude Code Review on 18 Mar 2026 as a managed GitHub PR-comment surface on top of Claude Code CLI. Pricing is $100/seat/mo Team plansrcplus pass-through Sonnet 4.5 tokens. A 10-dev team at 50 PRs/week and 250 LOC/PR burns ~$1,000 in seats + ~$54 in tokens per month.
@@ what shipped @@anthropic
1+A managed PR-comment surface that sits on top of Claude Code CLI. GitHub Checks + PR comments + reasoning trace.
2+Token meter passes through to the standard Anthropic API. Sonnet 4.5 list: $3/M input, $15/M outputsrc.
3-GitHub-only at launch. GitLab and Bitbucket not yet announced. Azure DevOps unclear.
@@ 10-dev burn model (50 PRs/wk, 250 LOC/PR) @@napkin math
1seat = $100 * 10 = $1,000/mo
2PRs/mo = 50 * 4.33 = 216.5
3tokens/PR = 250 * 24 input + 250 * 2 output
4token $/PR = (6000/1M)*$3 + (500/1M)*$15 = $0.018 + $0.0075 = $0.0255
5token total = $0.0255 * 216.5 = $5.52/mo
6+all-in: ~$1,006/mo at 10 devs
@@ how this changes the BYOK math @@downstream
1Claude Code Review IS the managed-BYOK hybrid. The seat just buys the surface; tokens go to Anthropic directly. So "BYOK vs Claude Code Review" is really "DIY surface vs Anthropic's surface" on the same token bill.
2For CodeRabbit, Greptile and Qodo the breakeven is unchanged because they bundle tokens. See /byok-breakeven.
@@ cross-links @@related
1/tool/claude-code-review - full file
2/vs/greptile-vs-claude-code-review - the head-to-head
3/byok-breakeven - downstream math