Claude Code: /code-review Now Uses 25% Fewer Tokens
Claude CodeClaude Code 2.1.196 makes the built-in /code-review workflow meaningfully more cost-efficient by merging five previously independent cleanup finder passes into a single consolidated step, reducing token usage by approximately 25% per review run. The improvement applies automatically to all /code-review invocations with no configuration required, making it an immediate win for teams running frequent code reviews across large codebases.
Key Takeaways
- The /code-review workflow now uses ~25% fewer tokens per run, achieved by merging five separate cleanup finder passes into one consolidated step: immediate cost savings with no user action required.
- No configuration is required: the efficiency improvement applies automatically to all
/code-review,/code-review medium, and/review <pr>invocations in 2.1.196 and later. - Teams running high-volume code reviews benefit most, as the savings compound across many pull requests and translate directly to lower API costs and faster review cycles.
- Fewer tokens also means fewer incomplete reviews on large diffs, since the reduced context footprint makes it less likely to hit token limits on substantial changesets.
- Anthropic is systematically optimizing built-in workflows for efficiency, following the /recap token optimization earlier in 2026 with this /code-review consolidation.
- The change ships alongside the background agent reliability overhaul in v2.1.196, signaling a release focused on practical developer experience improvements rather than net-new features.
Sources & Mentions
4 external resources covering this update
/code-review Gets 25% More Efficient
Anthropic shipped an internal restructuring of Claude Code's /code-review workflow in version 2.1.196: five previously separate cleanup finder passes have been merged into a single consolidated step. The result is a roughly 25% reduction in token usage on every /code-review invocation.
What Changed Internally
The /code-review workflow previously ran multiple independent analysis phases, each responsible for a distinct category of code quality checks: redundant logic, dead code, formatting drift, naming inconsistencies, and similar cleanup concerns. Each pass carried its own context loading and setup overhead. In 2.1.196, Anthropic consolidated five of these cleanup finders into a unified step, eliminating the overhead of repeated context initialization across phases.
The change is transparent from the user's perspective. Running /code-review, /code-review medium, or the updated /review <pr> command invokes the improved workflow automatically.
Why It Matters
Token costs compound quickly in high-volume development environments. For teams running /code-review across many pull requests daily, a 25% reduction translates directly to lower API costs and meaningfully faster review turnaround. Large diff reviews that previously risked partial analysis due to context length constraints are now more likely to complete fully in a single pass.
This efficiency improvement follows a broader pattern of Anthropic investing in the internal performance of Claude Code's built-in slash commands. The /recap command, shipped earlier in 2026, similarly reduced session-resumption token overhead by producing a summary rather than replaying full conversation history. The /code-review optimization continues that trajectory.