Krux
March 15, 2026
Anthropic's Code Review Triples PR Scrutiny Depth to 54%
Published: March 15, 2026 at 12:32 AM
Updated: March 15, 2026 at 12:32 AM
100-word summary
Anthropic just launched Code Review, a multi-agent feature that automatically audits pull requests before they ship. The payoff is real: before Code Review, only 16% of PRs got substantive feedback. With it, that jumps to 54%. Each review runs about 20 minutes and costs $15 to $25. Large PRs see the biggest gains—84% flag issues, averaging 7.5 problems per review. Small changes (under 50 lines) still surface findings 31% of the time. It's available now in research preview for Claude Code Teams and Enterprise users. The catch? You're paying per review, so teams will need to watch the analytics dashboard closely. Human approval still required—this doesn't greenlight code, it just...
What happened
Anthropic just launched Code Review, a multi-agent feature that automatically audits pull requests before they ship. The payoff is real: before Code Review, only 16% of PRs got substantive feedback. With it, that jumps to 54%. Each review runs about 20 minutes and costs $15 to $25. Large PRs see the biggest gains—84% flag issues, averaging 7.5 problems per review. Small changes (under 50 lines) still surface findings 31% of the time. It's available now in research preview for Claude Code Teams and Enterprise users. The catch? You're paying per review, so teams will need to watch the analytics dashboard closely.
Why it matters
Human approval still required—this doesn't greenlight code, it just makes sure more eyes (virtual ones) actually see it.