deepsec is an agent-powered security harness for finding vulnerabilities in large codebases. It is designed to run on the user’s own infrastructure, using coding agents to perform deep, project-aware security review rather than only basic pattern matching. The tool initializes a project-specific security context, scans code in batches, processes findings, and can optionally revalidate results to reduce false positives. For large repositories, it supports parallel work across worker machines and can resume interrupted scans without repeating completed work. It is aimed at serious security reviews where deeper model reasoning may be worth the higher compute and model cost. The project is especially relevant for teams that need to uncover long-standing vulnerabilities hidden inside complex application code.
Features
- Agent-powered vulnerability scanning
- Runs on user-controlled infrastructure
- Project-specific security context setup
- Parallel scanning for large repositories
- Interrupted run resume behavior
- Markdown findings export workflow