Claude Project Loader

Upload whole folders into a Claude Project as clean, path-named files. No name collisions. Nothing leaves your browser.

Claude Projects identify files by their name. So when you upload a codebase directly, files that share a name across different folders collide. An index.ts in your auth folder and an index.ts in your billing folder become the same file — and one silently overwrites the other. Claude ends up missing half your code, or treating two unrelated files as one. The bigger the project, the worse it gets. Claude Project Loader fixes this. It flattens each file's full path into its name before uploading, so src/auth/index.ts becomes src∕auth∕index.ts and src/billing/index.ts becomes src∕billing∕index.ts. Every file is preserved, every path stays readable, and nothing collides. WHAT IT DOES Path-named files — The full folder path becomes the filename, so same-named files across folders never overwrite each other. You can choose the separator style. Drop folders and files together — Drag a whole folder, a few loose config files, or any mix of both in a single drop. Add more before uploading; everything accumulates. Review as a folder tree — See the staged files as a collapsible directory tree. Toggle a whole folder in or out at once, rename a file inline, or remove files and folders from the selection entirely. Token & capacity awareness — See the tokens each file (and the whole upload) will consume, and how the upload lands against the project's capacity — "42% → 88%" — before you commit, so you never silently overfill a project. Duplicate detection — Reads the files already in the open project and flags any staged file whose name already exists there, so you don't upload a stale second copy that confuses Claude. Skips the noise — Honors your .gitignore, a repo-local .claudeignore (same syntax, shareable through version control), and a built-in ignore set (node_modules, build output, lockfiles, binaries, and more). All toggleable. Flags low-value files — Surfaces empty, comment-only, and files that are real but unhelpful to Claude (minified bundles, source maps, vendored or generated code, bulk data) so you can drop them in one click. Protects secrets — Detects files that likely contain secrets — .env files, private keys, credentials — and excludes them by default, with a clear warning. You always keep the final say. Manage the project — See what's already in the project and delete files from it directly, without leaving the panel. Flexible output — Add files straight into the project, save them to a folder to drag in yourself, or merge everything into a single document with a file tree and clear per-file headers. (No zips — Claude Projects can't expand them.) Optional cleanup — Conservatively strip comments and blank lines to reduce size, only where it's provably safe for the file's language. PRIVACY FIRST Everything runs locally in your browser. Your files are never uploaded to any server, never stored, and never sent anywhere except the upload into your own Claude Project that you trigger. There is no analytics, no telemetry, and no tracking. The only thing the extension saves is a single local number used to make the capacity estimate more accurate — it never leaves your browser. The extension requests two permissions (activeTab and storage) and runs only on claude.ai. The processing engine contains no network code at all — a claim you can verify yourself, because the whole project is open source. WHO IT'S FOR Developers and teams who want to give Claude real context about a codebase — the whole repo, organized and clean — without fighting filename collisions or manually renaming files one at a time. OPEN SOURCE Free and MIT-licensed. Source, issues, and contributions: https://github.com/ramanaditya/claude-project-loader