What AI coding assistants does DeepSweep support?

DeepSweep supports all major AI coding assistants including Cursor, GitHub Copilot, Microsoft Copilot, Claude Code, Windsurf, Cody, Tabnine, Amazon CodeWhisperer, and any tool that generates code.

Does DeepSweep work with cloud platforms?

Yes, DeepSweep integrates with AWS, Azure, Google Cloud Platform, Vercel, Netlify, and other cloud providers. It can run in any CI/CD pipeline including GitHub Actions, GitLab CI, CircleCI, and Jenkins.

Is my code sent to external servers?

No, DeepSweep runs locally by default. Your code never leaves your machine. For team features, only metadata (not source code) is transmitted over encrypted connections.

Changelog | DeepSweep — Authorize Every Action Your AI Agents Take

Added

Your coding agent is recognized by name on the very first review. The Agent Environment Review now identifies the agent you actually use — including Cursor, Claude Code, Windsurf, GitHub Copilot, and OpenHands — directly from your local setup, instantly and without waiting on any account or network. Your review's Agent Identity block names the real agent from review one.

Added

See exactly which agent identity your review covers — agent, owner, and session — at a glance. The Agent Environment Review now shows an Agent Identity block right above the capability list, so every review is clearly tied to the agent it describes. Free for everyone, on every tier.
Your agent is recognized automatically — on your very first review. Cursor (including the newest versions), Claude Code, Windsurf, GitHub Copilot, and OpenHands are identified from your editor and the agent files already in your workspace, so the review starts with the right identity instead of a generic one.
Pro: export your authorization policy as a portable artifact your other tools and pipelines can consume. A new "Export Authorization Policy" button in the review panel (also available as "DeepSweep: Export Authorization Policy" in the Command Palette) produces a clean JSON policy document — your rules, decisions, and agent identity in one portable file.

Fixed

Windows reviews are now as accurate as macOS and Linux. On Windows, some security checks silently skipped files because of path differences — reviews could miss findings or report different results than on other machines. Every check now applies to the same files on every platform.
Saving a config file no longer runs the same review twice.
Every panel is readable on compact displays. On smaller screens (including 800x600), the welcome screen no longer cuts off its heading, long file paths wrap instead of forcing sideways scrolling, and buttons, filters, and stat rows reflow to fit narrow side panels.

Changed

Steadier performance in long sessions. Review results are cached more efficiently while you type, keeping memory use flat during extended editing.

Stay Protected