Anthropic has extended Claude Code's code review function with an effort-level control. As the company announced via its official bot, reviews are rewritten at each effort level – this is not a simple filtering mechanism but rather a difference in analysis depth.
At a glance
- Low-effort mode delivers better findings than other code review tools according to Anthropic, while consuming only a fraction of the token cost
- High-effort mode delivers significantly higher recall rates for deeper analysis, according to Anthropic
- Users can choose their own tradeoff between cost and analysis depth
- The feature is available via the
/code-reviewcommand in Claude Code
What this means
The effort levels address a classic tension in AI-assisted code analysis: thoroughness versus cost efficiency. For German development teams, the low-effort setting could be interesting if it truly delivers better results than established tools at lower token consumption – that would fundamentally change the cost calculation for AI-driven reviews. What remains unclear is how the two modes differ in practice and which error classes each captures.
Sources
Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.
