Why AI coding tools consume so much disk space
AI development tools behave like editors, package managers, local databases, and agent runtimes at the same time. A single workflow can create model caches, extension data, logs, workspace state, backups, generated output, and temporary files across several Library folders.
Those files are difficult to judge in Finder because they are spread across application support, caches, logs, and project directories. DiskTidy groups the metadata by tool and purpose so you can distinguish active data from re-creatable or orphaned residue.
What DiskTidy detects
- Caches and logs created by supported AI editors and command-line tools
- Workspace state, extension storage, backups, and known application data roots
- Orphaned roots from supported editors that are no longer installed
- Active AI processes and their current memory use
- Disk pressure, available memory, thermal state, and disk health signals
Safe cleanup with evidence first
DiskTidy lists the exact candidate, category, path, and estimated recoverable space before a cleanup starts. Running tools remain protected, and project directories are excluded from automatic cleanup.
Safe Clean is designed for routine maintenance of re-creatable data. Full Clean is intentionally stricter: it re-checks that the supported editor is absent and inactive, limits deletion to known allowlisted roots, and requires a second confirmation.
Private by architecture
Scanning happens locally on your Mac. DiskTidy works from metadata such as paths, file names, sizes, dates, file types, and access status. It does not send project contents, prompts, sessions, or command history to a remote analysis service.
Frequently asked questions
Which AI coding tools can DiskTidy inspect?
DiskTidy recognizes data associated with Cursor, Claude Code, Codex, Windsurf, Cline, Copilot, Replit, Lovable, and other supported AI development workflows.
Does DiskTidy read prompts or source code?
No. AI Tool Health uses file-system metadata and system signals. It does not upload or display source code, prompt text, session contents, configuration contents, or process command-line arguments.
What is the difference between Safe Clean and Full Clean?
Safe Clean targets re-creatable caches and logs while protecting running tools. Full Clean is a separate, two-step action for exact allowlisted roots belonging to supported editors that are no longer installed or active.