DeepSweep: AI Agent Security & Code Review

deepsweep-ai.deepsweep · by deepsweep-ai·Open VSX Verified

Security review for AI coding agents: AI agent security, vulnerability detection and secrets review, locally. Your AI agents have shell access, repo write permissions, and your credentials. Do you know what they can do without asking you? DeepSweep reads your agents' settings and MCP configs locally. It shows agents allowed to run any shell command, or to push, deploy or publish, without asking you. It shows access beyond the project, such as the whole disk, your home folder or no sandbox. It flags MCP servers that are not pinned to a version or that reach a remote server with no sign-in, and credentials written in plain text into agent configs. The agent environment review runs in under a second. DeepSweep also reviews the code your agent writes as it lands for security vulnerabilities and secrets, and records what it found in a local report you can export. Runs in Antigravity, Trae, Devin Desktop (formerly Windsurf), Kiro and VS Code, and in Cursor when installed from the VSIX, alongside Claude Code, OpenAI Codex, GitHub Copilot, Gemini Code Assist, Cline, Tabnine, Amazon Q Developer and Aider. Run your first review free. No account required.