A new solution for code review from Anthropic
Recently, the startup introduced a new feature within Claude Code — a multi-agent tool for reviewing source code. It easily integrates with GitHub and can analyze pull requests, leaving comments directly in the diff, as well as generating a comprehensive report with key findings.
What’s particularly interesting is that multiple agents work on a single PR simultaneously, each examining the changes from their own perspective. This approach allows for a more diverse and in-depth analysis.
The company has been testing this system on its internal projects for several months. The results are impressive: the share of PRs with meaningful comments increased from 16% to 54%. An important detail is that less than 1% of the evaluations were recognized by engineers as incorrect. For large PRs (around 1000 lines), the bot found at least one issue in 84% of cases, typically identifying 7–8 deficiencies per change.
Overall, everything looks very promising. However, each such analysis costs approximately $15 to $25, which is not insignificant.
Honestly, this tool seems especially useful for large PRs, often generated automatically or containing many changes. In such cases, the costs may be justified, but for smaller tasks, the price is unlikely to be worthwhile.
Created with n8n:
https://cutt.ly/n8n
Created with syllaby:
https://cutt.ly/syllaby