Cursor IDE
Built on the VS Code and Electron stack, Cursor IDE stores workspace state, extension metadata, IndexedDB content, SQLite databases, and Chromium browser data under its application support folder so windows, panels, authentication flows, and AI-assisted features can reopen quickly. Over time those SQLite files accumulate rollback journals, WAL files, fragmented pages, and stale webview storage that can survive crashes or extension changes. Kudu clears disposable Cursor database and browser-cache data, including SQLite support files and stale webview storage, without touching your source files, settings, accounts, or saved credentials.
Why clean Cursor IDE?
- Crash-left SQLite WAL or journal files can make Cursor reopen with missing recent UI state, repeated restore prompts, or panels that never finish loading
- Fragmented SQLite databases grow far larger than their live data, so Cursor uses more disk and searches workspace history or recent items more sluggishly until the database is rebuilt
- VACUUM rewrites SQLite databases to remove free pages without deleting live rows, which helps when long-term extension churn leaves internal Cursor databases bloated
- Stale IndexedDB and webview storage from an updated extension or chat panel can cause blank sidebars, stuck sign-in popups, or AI panes that spin indefinitely
- Corrupted workspace storage can make the same project reopen with lost panel layout, duplicated tabs, or prompts to restore editors every launch
- Old browser-session data inside Cursor's Electron layer can trap outdated authentication state, showing repeated login loops or embedded pages that fail to load correctly
Cache paths Kudu targets
These databases are vacuumed (compacted) — no data is deleted. SQLite VACUUM reclaims wasted space left over from normal usage.
Windows
%AppData%/Cursor |
macOS
~/Library/Application Support/Cursor |
Linux
~/.config/Cursor |
Common questions about Cursor IDE
Download Kudu and reclaim your disk space.
Available on Windows, macOS, and Linux. No account required, no feature gates, no telemetry without consent. All cleaning targets are open source and community-auditable.