Tabnine's new Code Review Agent supports enforcing organization-specific rules

Tabnine recently launched its Code Review Agent under a rather ambitious promise: unlike static code analysis tools, and AI-powered code review products currently on the market, Tabnine's Code Review Agent can be trained on a team's standards, best practices, and policies, and remain aware of these custom specifications when reviewing code. The Code Review Agent integrates directly into IDEs and pull requests, providing real-time feedback and suggested fixes. This automation allows developers to focus on higher-order analysis rather than routine compliance checks.

According to Tabnine, users need only provide the Code Review Agent with the guidelines it must adhere to using plain English, for instance, by sharing documentation or a wiki as sources. There is also an option to provide access to selected code repositories instead of documents, where Tabnine will look for recurring PR comments and other relevant code patterns. In both cases, the extracted information will be automatically transformed into rules for code validation that the Code Revie Agent will follow. The rules are also documented in natural language, so administrators can go over them, and decide to accept, reject, or edit them so they better conform to the organization's practices.

In parallel, Tabnine has enabled users to complement their customized rules with predefined rules cautiously designed for each supported language based on widely accepted best practices documents and consultations with language-specific experts. Their varying specificity ensures that, upon adoption, these rules allow the Code Review Agent to validate the code's high-level structure and specific functions.

The Tabnine Code Review Agent is in private preview at the moment, and Tabnine Enterprise customers can contact support to request early access. To learn more about the product, those interested can review Tabnine's documentation or join the Tabnine Live session on November 7 at 8 am PT.