AI coding tools and IDEs are using a lot of disk space
A few active projects can leave behind IDE caches, package downloads, temporary agent folders, old worktrees, and several complete sets of build output. They may all be under different parts of your profile and source tree.
There are several different kinds of developer storage
An IDE cache is usually generated data that the application can recreate. A Rust target folder or a Delphi DCU folder is project build output. A Codex or Claude scratchpad may contain useful session work. A Kiro conversation or Git worktree can contain the only local copy of something important.
Those folders may sit beside each other in a disk scan, but they should not be cleaned the same way. Start by identifying the owner and the kind of data before deciding whether it is safe to remove.
Start with the Overview pane
After scanning the development drive, check SizeTrend's Overview pane. It can recognize generated cache folders for Visual Studio Code, Cursor, Kiro, and Claude Desktop. It also groups Codex temporary artifacts, Claude session scratchpads, Kiro session history, and older Kiro diagnostic logs separately.
The names matter. SizeTrend does not label an entire application profile as cache. Cursor chat transcripts, Kiro conversations, account state, extensions, and Claude Code installations are deliberately excluded from the simple cache recommendations.
Visual Studio, VS Code, and Delphi projects
For recent Visual Studio projects, SizeTrend looks for generated output below the project ownership root without evaluating MSBuild. Delphi projects are checked for DCU, executable, package, library, and intermediate output folders. Output directed to an external runtime folder is not guessed or treated as project cleanup.
Recent VS Code workspaces can also expose large Node dependencies and build folders, Python virtual environments and test caches, Maven or Gradle output, and configured CMake build trees. Reinstalling dependencies or rebuilding a large project can take time, so review the project before running its normal Clean command.
Rust and Cargo target folders
Rust build output can grow quickly when a workspace has several profiles, feature combinations, or large dependency trees. SizeTrend recognizes a target folder only when it is beside an indexed, non-reparse Cargo.toml manifest. It does not guess configured target directories elsewhere on the drive.
When the project is still in use, prefer cargo clean from the package or workspace root instead of deleting individual files. Remember that the next build may need to compile the complete dependency tree again.
npm, NuGet, pip, and downloaded PDB caches
Package managers keep their own shared download caches outside a project. SizeTrend recognizes the npm package cache, NuGet's global package, HTTP, and plugin caches, and the pip cache. Its Clean action asks npm, dotnet or NuGet, and pip to clear their managed data when those tools are available. Installed projects and environments are left alone, but the next restore may download the packages again.
Visual Studio, WinDbg, and diagnostic tools can also accumulate downloaded PDB and image files. SizeTrend cleans only configured or standard symbol-cache locations that have the expected SymSrv marker. A folder that merely looks like a symbol store is shown for review instead, because it could contain private or irreplaceable symbols.
Close the tool and use its own cleanup path
Quit Cursor, Kiro, Claude Desktop, or VS Code before cleaning a generated cache. Confirm that Codex and Claude temporary paths do not belong to an active task. Use Visual Studio or Delphi Clean for active projects, and remove linked Git worktrees with git worktree remove.
SizeTrend helps separate these categories and shows the oldest or largest project paths. It never assumes that every folder with an AI tool or IDE name is disposable.
Find the generated data without treating the whole project as junk.
SizeTrend groups AI-tool storage, IDE caches, worktrees, and build output by their actual role.