--- title: MangoDisk: How to find and delete junk on macOS/Windows risk-free url: https://blog.krasovskiy.team/en/mangodisk-how-to-find-and-delete-junk-on-macos-windows-risk-free/ date: 2026-08-21 lang: en source: blog.krasovskiy.team --- # MangoDisk: How to find and delete junk on macOS/Windows risk-free Finding on the hard disk something that steals tens of gigabytes is not an easy task: Xcode or Docker caches, Node.js artifacts after another ``npm install'', or local copies of AI models that have not been needed for a long time. MangoDisk solves this without the risk of accidentally deleting something critical - the tool scans the disk in read-only mode, and the deletion requires manual confirmation, relying on the rules from the official documentation of Apple, Microsoft and other vendors. 938 stars on GitHub, updated on August 20th this year, and the ability to test the potential cleanup result with a single command: `mangodisk clean --apply --selection all --dry-run`. ## What is MangoDisk and why is it needed MangoDisk is a disk space analysis and cleanup tool that works on macOS and Windows. It is designed for those who want to free up disk space without the risk of accidentally deleting important files. Especially useful for developers: Finds and removes garbage left behind by IDEs, package managers, Node.js, Rust, Gradle, Swift, Python artifacts, Xcode leftovers, Docker, and even [local AI models](https://blog.krasovskiy.team/en/echobird-a-universal-tool-for-quickly-managing-ai-models/). Unlike aggressive "cleaners", MangoDisk works in read-only mode by default - files can be deleted only after confirmation, and cleaning rules are based on the official documentation of vendors. The interface shows the visualization of disks in the form of a treemap, allows you to search for duplicates by content, has a built-in uninstaller and CLI for automation. For example, to test cleaning without deleting, it is enough to execute the command: `mangodisk clean --apply --selection all --dry-run`. The tool is written in Rust, uses Vue 3 and Tauri 2, has 938 stars on GitHub, and is regularly updated — the latest commit is dated August 2026. ## What types of garbage does MangoDisk find MangoDisk finds junk that accumulates invisibly, from IDE caches to development artifacts. Here is what it analyzes: - **IDE and editor caches**: leftovers from VS Code, JetBrains (IntelliJ, PyCharm), Xcode, Sublime Text. For example, Xcode stores tens of gigabytes of temporary files after every build. - **Development artifacts**: folders `node_modules`, `target` (Rust), `build` (Gradle), `__pycache__` (Python), `.swiftpm` (Swift). These directories are often duplicated in each project. - **Docker**: abandoned images, containers, volumes and networks. A single undeleted image can take up 5-10 GB. - **Local AI models**: files from `~/.cache/huggingface`, `ollama/models` or `stable-diffusion`. The Llama-3 model weighs 16 GB — and often remains after experiments. - **Package managers**: Homebrew, npm, pip, Cargo, Composer caches. For example, `npm cache` can grow to hundreds of megabytes. - **Duplicate files**: files that are identical in content, even if they have different names. The tool compares hashes, not just size. MangoDisk does not guess what can be removed - it relies on the official documentation of the vendors (for example, Docker Prune or npm cache clean). By default, it works in read mode: deletes only after your confirmation. In test mode (`--dry-run`) will show how much space will be freed without risk. ## Security and features of working with MangoDisk MangoDisk puts security first - and it is felt from the first seconds. By default, the tool works in **read mode**: no file will disappear without your knowledge. Before uninstalling, the system requires **confirmation** - even if you've selected hundreds of gigabytes of Xcode caches or Docker artifacts. This protects against accidental losses, especially when working with critical projects. The **treemap** visualization shows exactly where the space is "eaten up": large blocks are large files or directories, colors are categories (for example, IDE caches or AI models). I clicked on the block — and you can see the exact path. It's faster than digging around in the terminal with `du -sh`. For those who want to automate the routine, there is a **CLI**. For example, the command `mangodisk clean --apply --selection all --dry-run` will show how much space will be freed up without actually deleting. And **duplicate search** by content (not just by name) will help you find duplicates even among different directories. All this is based on rules from the official documentation of vendors, not "aggressive" algorithms. Stack? Rust for speed, Vue 3 + TypeScript for frontend, Tauri 2 for cross-platform. 938 stars on GitHub is not an accident: the tool solves specific problems of developers, not just "cleans the disk". ## How MangoDisk helps AI and SEO professionals AI and SEO experts face the same problem every day: disk space is running out without notice, and manual cleanup is like looking for a needle in a haystack. MangoDisk solves this systemically. For AI engineers, the tool finds and removes local copies of models (such as unused versions of Stable Diffusion or Llama), Docker container artifacts, and temporary files of frameworks like PyTorch. Instead of manually combing through folders, just run a scan and see that 15GB is occupied by outdated model checkpoints and another 8GB by the CUDA cache. The CLI command `mangodisk clean --apply --selection ai-models` automates the process, freeing up space without the risk of deleting critical data. SEO specialists also benefit. Large projects with hundreds of pages generate tons of temporary files: browser caches, local copies of sites for testing, archived versions of content. MangoDisk analyzes caches of IDEs (WebStorm, VS Code), package managers (npm, pip) and Docker images that grow to tens of gigabytes. For example, the Node.js cache alone in a large project can "eat" 5-7 GB. The tool shows this in a visual treemap, where you can immediately see what exactly is taking up space. And the function of searching for duplicates by content (not by name!) helps to get rid of copies of images or documents that have been accumulating for years. The result is faster tools, less time spent on maintenance, and more resources for analytics. Automation is another trump card. Instead of a weekly manual cleanup, you can set up a regular run via the CLI or integrate MangoDisk into your CI/CD pipeline. For example, for AI teams: after training the model, automatically delete temporary files, leaving only the final artifacts. For SEO, clear the cache after every major site update. 938 stars on GitHub and an update in August 2026 confirm: the tool does not just work, but evolves along with user requirements.