OpenAnt uses large language models to proactively discover security vulnerabilities in source code across multiple programming languages (Go, Python, JavaScript/TypeScript, C/C++, PHP, Ruby). It operates in two stages: first detecting potential vulnerabilities, then attempting to attack them to verify which findings are genuine. The tool is designed to minimize both false positives and false negatives by combining LLM-based detection with automated attack validation, helping defenders identify real security flaws in their repositories.