Tabnine AI agent is designed to enforce a development team’s best practices and standards throughout the software development process, using natural language rules.
Tabnine, maker of the Tabnine AI coding assistant, has unveiled the Tabnine Code Review Agent, an AI-powered software validation agent that helps users to produce higher quality, more secure code.
Now available in a private preview for Tabnine enterprise customers, the Tabnine Code Review Agent is designed to enforce a development team’s best practices and standards throughout the software development process. A demo of the tool can be found on YouTube. With the Tabnine Code Review Agent, users can codify institutional knowledge, corporate policies, and software development standards. Guidance offered by the tool builds on Tabnine’s approach to AI code generation through awareness and understanding of locally available code, data in the IDE, and a company’s software repository, Tabnine said.
“AI in software development is about more than generating code. Its greatest power might be in helping to improve quality, security, and compliance of code in real time,” Tabnine president Peter Guagenti said in a statement. “By reviewing code at the pull request and ensuring that the code presented matches each team’s unique expectations, we are saving engineering teams significant time and effort while applying a level of rigor to the automation of code review that was never possible with static code analysis,” Guagenti said. Tabnine offers an approach to personalization that allows agents to behave like an onboarded member of an engineering team familiar with a company’s ways of working, he added.
Users of Tabnine Code Review Agent can provide specific parameters they would like to see their code comply with via plain language, with no complex setup needed, Tabnine said. Tabnine then converts this knowledge into a set of rules. Tabnine also offers predefined rules any team can activate. When developers create a pull request, the Tabnine Code Review Agent checks the code in the pull request against the rules established by their team. If any aspect of the code does not conform to those rules, the agent flags it to the code reviewer, offering guidance on the issue and suggest edits to fix it. All of the rules are in plain English, making them easy to review and maintain over time, Tabnine said.
