Anthropic introduced an artificial intelligence capable of finding errors in code generated by machine learning

Anthropic’s new Code Review tool will help you fix code bugs faster

As the automation of software development grows, bugs inevitably appear. Artificial intelligence not only generates them but also knows how to find and eliminate them. That is why Anthropic introduced Code Review—a new service for checking code available within the Claude Code platform.

What is Claude Code?
Claude Code is an AI agent that writes program code. Thousands of developers around the world actively use it today. Usually, many changes and fixes are required before a product release.

Code Review is designed to speed up this process: code checks are performed automatically or with minimal human involvement.

How the tool works
1. GitHub integration

Code Review connects to a repository, analyzes proposed changes, and generates comments.

2. Multiple agents

Several AI agents work in parallel, each reviewing code under different criteria (logic, security, style).

3. Final summary

One agent consolidates all remarks: removes duplicates and highlights the most important ones.

4. Color-coded problem marking

* Red* – critical errors that require immediate correction;
* Yellow* – potential issues worth paying attention to;
* Purple* – mistakes in previously generated code.

5. Cybersecurity

The tool performs an initial security audit and allows the user to set up additional checks.

Cost
Anthropic uses a pay‑per‑use model. Analyzing one snippet of code costs between $15 and $25. This can become a significant expense if a project actively uses AI for development.

Why it matters
According to Anthropic representatives, the speed of code creation should not compromise its quality. Therefore automatic error‑checking tools are becoming increasingly in demand in the software industry. Code Review is already available as an experimental solution in Claude for Teams and Claude for Enterprise.