Anthropic just rolled out Code Review, a new tool built into Claude Code that automatically checks pull requests for bugs before they hit the codebase. The launch, reported by TechCrunch AI, targets enterprise teams drowning in AI-generated code that needs human oversight.
The timing is sharp. As “vibe coding” floods engineering teams with machine-written pull requests, someone still needs to catch the bugs. Anthropic is betting that an AI reviewer can handle that bottleneck faster than human engineers alone.
“We’ve seen a lot of growth in Claude Code, especially within the enterprise, and one of the questions that we keep getting from enterprise leaders is: Now that Claude Code is putting up a bunch of pull requests, how do I make sure that those get reviewed in an efficient manner?” Cat Wu, Anthropic’s head of product, told TechCrunch.
What Code Review Actually Does
Once enabled, the tool integrates with GitHub and automatically analyzes every pull request on a team. Here’s the breakdown:
- Logic over style. Anthropic deliberately skipped nitpicky formatting feedback. The tool focuses on logical errors, the kind of bugs that actually break things.
- Step-by-step reasoning. Each flagged issue includes an explanation of what’s wrong, why it matters, and how to fix it.
- Color-coded severity. Red for critical issues, yellow for potential problems worth a look, purple for issues tied to preexisting code or historical bugs.
- Multi-agent architecture. Multiple AI agents examine the codebase in parallel from different angles. A final agent aggregates findings, removes duplicates, and ranks by priority.
- Light security analysis. Basic security checks are included, though Anthropic points to its separate Claude Code Security product for deeper scans.
- Customizable rules. Engineering leads can add checks based on internal best practices.
Wu made a deliberate design choice here. “A lot of developers have seen AI automated feedback before, and they get annoyed when it’s not immediately actionable,” she said. Focusing purely on logic errors means the tool catches high-priority problems without generating noise.
Who Gets It and What It Costs
- 📍 Availability: Research preview for Claude for Teams and Claude for Enterprise customers. Not available on free or Pro plans yet.
- 💰 Pricing: Token-based, similar to other AI services. Wu estimated each review costs $15 to $25 on average, depending on code complexity. That’s a premium price point, and Anthropic isn’t shy about it.
- 🏢 Target users: Large-scale enterprise teams. Wu named Uber, Salesforce, and Accenture as the kind of companies this is built for.
Developer leads can turn on Code Review to run by default for every engineer on the team, meaning individual devs don’t need to opt in.
Why This Matters Now
Anthropic’s enterprise business is on a tear. Claude Code’s run-rate revenue has surpassed $2.5 billion since launch, according to TechCrunch AI, and enterprise subscriptions have quadrupled since the start of 2026. Code Review is a natural expansion: the same tool that generates the code now helps review it.
What stands out here is Anthropic positioning itself on both sides of the AI coding equation. Claude Code creates the pull requests; Code Review checks them. That’s a compelling loop for enterprise buyers who want speed without sacrificing quality.
The $15-$25 per review price tag will raise eyebrows for smaller teams, but for enterprises already paying for Claude Code at scale, it’s a rounding error compared to the engineering hours saved. The real question is accuracy. If the tool generates too many false positives or misses critical bugs, developers will ignore it just like they ignore linters.
Anthropic is betting that won’t happen. “We’re hopeful that with this, we’ll enable enterprises to build faster than they ever could before, and with much fewer bugs than they ever had before,” Wu said.
More details are available in the full TechCrunch AI report.