lintsinghua/DeepAudit

DeepAudit is a multi-agent code vulnerability detection system that automates security auditing through collaborative AI agents and sandbox-based proof-of-concept verification.

The tool addresses the challenge of finding code vulnerabilities at scale by deploying multiple AI agents that work together to analyze code, identify potential security issues, and validate findings through automated sandbox testing. It supports private deployment via Ollama, integrates with version control platforms including GitHub, GitLab, and Gitea, and generates security reports in multiple formats. The system is designed for accessibility, allowing users to begin analysis with minimal setup while providing real-time visibility into agent reasoning and execution.

DeepAudit suits teams seeking to reduce security audit costs and complexity without requiring specialized expertise. It works well for projects that need continuous vulnerability scanning across multiple repositories and benefit from automated report generation. The tool distinguishes itself through its multi-agent orchestration approach and integrated sandbox verification rather than relying solely on static analysis patterns.

The project maintains a substantial adopter base, with nearly all open issues originating from external users rather than the core team. Maintainers respond to new issues and pull requests within hours. Development activity centers on enhancements and bug fixes, indicating active refinement of existing capabilities.