Selected customers can now evaluate CodeMender through Google’s Gemini platform, but they cannot yet deploy its vulnerability scans or proposed patches in commercial software.
Google began allowing a limited number of customers to test CodeMender, an AI agent that finds, verifies, and proposes fixes for security flaws in software code.
The announcement does not introduce a new AI security capability. Google described CodeMender’s main functions in late May when it announced AI Threat Defense, a broader platform that combines several Google security products. At the time, Google said customers would test CodeMender through its Gemini Enterprise Agent Platform.
What changed this week is that CodeMender has formally entered public preview. Customers can now request access through their Google Cloud sales teams and test the agent as a managed service through the Gemini platform.
Despite the “public preview” label, access is limited. Google’s documentation says CodeMender is available only to selected customers and may not be used for commercial or production purposes. The company says the preview is intended solely for testing and evaluation.
CodeMender focuses on software code
CodeMender scans a company’s software for vulnerabilities, builds test attacks to determine whether suspected flaws can be exploited, and generates proposed patches for confirmed problems.
The agent runs those tests in an isolated environment controlled by the customer. It then tests the proposed patch to check whether the change fixes the vulnerability without disrupting other parts of the software.
Developers manually review and approve patches before they are added to a company’s code repository.
Google’s documentation warns that CodeMender may modify files or execute commands on a customer’s system and recommends running it in an isolated environment.
The preview supports C, C++, Go, Java, JavaScript, Python, Ruby, Rust and TypeScript. Customers can choose among several Google Gemini models depending on the type of scan they want to run. Google said CodeMender will support advanced models from other companies later this year.
CodeMender: One piece of AI Threat Defense
CodeMender is a specialized code-security agent. AI Threat Defense is the broader security platform that combines CodeMender with Wiz, Mandiant, and other Google security products.
Within that platform, CodeMender handles software code. Wiz adds information about how applications and cloud resources are deployed and exposed, allowing security teams to determine which vulnerabilities present the greatest risk. Other parts of AI Threat Defense monitor systems for attacks and support incident response.
Google’s latest announcement says Wiz will eventually be able to direct CodeMender to scan software and propose fixes using that broader security information. CodeMender scanning through the Wiz integration is still listed as “coming soon.”
Customers can test CodeMender directly through the Gemini Enterprise Agent Platform without using the full AI Threat Defense platform.
Google has not announced general availability, pricing, or a timeline for allowing companies to use CodeMender in commercial systems.
