Context Lens for GitHub

Context Lens turns a GitHub repository or folder into Markdown you can paste into an AI chat assistant, and shows how many tokens it will use while you choose what goes in. HOW IT WORKS Open a repository or folder page on GitHub and press "Prepare for AI". The extension downloads the repository as a single zip, counts tokens for every file, and shows a folder tree with a token count on every row and a gauge against your model's context window (32K, 128K, 200K or 1M). WHAT YOU CAN DO • Exclude tests, docs or examples with one checkbox. In many repositories, tests are most of the tokens. • Lean mode: drop comment-only lines and extra blank lines. • Skeleton mode (JavaScript, TypeScript, Python, Go): keep signatures, types and docstrings, and drop function bodies. Measured on 7 real repositories: 35% to 70% fewer tokens, with no syntax errors in the output. • Fit to window: one click proposes what to drop to fit your chosen context window. It drops non-code folders first, never cuts code below half, and tells you plainly when something still does not fit. • Tick any folder or file on or off, then copy the Markdown or download it as a .md file. The output starts with a file tree, and files shown as skeletons are marked "(signatures only)". • Works on public repositories with no setup. For private repositories, add a read-only GitHub token in settings. PRIVACY Everything runs in your browser. There is no server, no analytics and no ads. Your optional token stays on your device and is sent only to api.github.com. The parser used for Skeleton mode is bundled with the extension, and no remote code is loaded. GOOD TO KNOW • Token counts are estimates made with the OpenAI o200k tokenizer. Other models count a little differently. • Repositories whose zip is over 60 MB are not supported. If a repository has more than about 25 MB of text, the first 25 MB is loaded and you are told so. • Binary files, lock files, dependency and build folders, translation files and files over 100 KB are skipped. • Branch names that contain a slash (for example feature/x) are not supported on /tree/ pages.