Hybrid AI-Static Analysis Tool BrassCoders Aims to Sharpen Python Code Security Reviews

BrassCoders is a code-review tool that runs 12 static-analysis scanners across Python codebases before passing structured findings to an AI assistant, combining the precision of deterministic rules with the contextual reasoning of large language models. The tool generates a severity-sorted file containing file paths, line numbers, and remediation guidance, giving the AI a defined starting point rather than requiring it to scan code from scratch. Two 2025 research papers support this hybrid approach: one found LLMs outperformed static tools in recall but struggled with exact line-level accuracy, while another reported F1 scores above 0.91 when static-analyzer output was fed to an LLM for adjudication. In an internal benchmark, BrassCoders scanned 15 AI-generated Python files, surfaced 53 findings including 16 critical or high-severity issues, and confirmed real vulnerabilities in 9 of the 15 files. The developers argue the pre-pass approach becomes especially valuable in large codebases, where filtering thousands of raw scanner findings down to a manageable set makes AI-assisted review more practical and consistent.