TokenSaver - Image Optimizer 作者： TokenSaver
Automatically shrinks images before you upload them to ChatGPT or Claude to reduce token usage. 100% local: no uploads, no tracking.
扩展元数据
关于此扩展
TokenSaver – Image Optimizer
Cut the token cost of screenshots and images in AI chat. TokenSaver automatically optimizes every image you attach in ChatGPT, Claude and Cursor's web agents, before it leaves your browser. Fewer pixels can mean fewer tokens, so your usage limits last longer.
Features:
- Automatic: works with the file picker, drag & drop and pasted screenshots
- Site-aware sizing: each service counts image tokens differently, and TokenSaver sizes images accordingly
- Real pixels, no OCR: the model always receives your actual image
- Transparent: a notice shows the estimated tokens before → after for every image
- Three levels: Auto (1024 px), Sharp (1280 px), Max saving (768 px)
- Safe: if an image can't be improved, it is sent unchanged
Supported sites: chatgpt.com, claude.ai, cursor.com/agents. Requires Firefox 120 or later.
Privacy: 100% local processing. No uploads to any server, no analytics, no tracking. The only permission used to store data is "storage", for your settings.
Token numbers are estimates based on public documentation, not exact billing. Reducing pixels can affect tasks that need very fine detail; use the Sharp level or turn the extension off in that case.
TokenSaver is independently developed and is not affiliated with or endorsed by OpenAI, Anthropic, Cursor or Mozilla.
Cut the token cost of screenshots and images in AI chat. TokenSaver automatically optimizes every image you attach in ChatGPT, Claude and Cursor's web agents, before it leaves your browser. Fewer pixels can mean fewer tokens, so your usage limits last longer.
Features:
- Automatic: works with the file picker, drag & drop and pasted screenshots
- Site-aware sizing: each service counts image tokens differently, and TokenSaver sizes images accordingly
- Real pixels, no OCR: the model always receives your actual image
- Transparent: a notice shows the estimated tokens before → after for every image
- Three levels: Auto (1024 px), Sharp (1280 px), Max saving (768 px)
- Safe: if an image can't be improved, it is sent unchanged
Supported sites: chatgpt.com, claude.ai, cursor.com/agents. Requires Firefox 120 or later.
Privacy: 100% local processing. No uploads to any server, no analytics, no tracking. The only permission used to store data is "storage", for your settings.
Token numbers are estimates based on public documentation, not exact billing. Reducing pixels can affect tasks that need very fine detail; use the Sharp level or turn the extension off in that case.
TokenSaver is independently developed and is not affiliated with or endorsed by OpenAI, Anthropic, Cursor or Mozilla.
评分 0（1 位用户）
权限与数据
更多信息
Supported sites: chatgpt.com, claude.ai, cursor.com/agents. Permissions: "storage" (saves the on/off switch and compression level) and host access to the supported sites, used only to process images the user attaches.
The extension is independently developed and is not affiliated with or endorsed by OpenAI, Anthropic, Cursor, Mozilla, or any other provider.
Source code: https://github.com/Taha2hussein/TokenShrink
Build: Node.js 18+, run "npm install" then "npm run build". The output is written to dist/.