TokenTrim

Compress your prompts before sending. Same answer, fewer tokens. Runs locally.

You write prompts like emails. "Could you please take a look at this and help me understand…" The AI doesn't need any of that, but you pay for every word, in tokens, in speed, and in how fast you hit your usage limit. TokenTrim fixes it without changing how you work. Press Send as usual. On the way out, it cuts the filler, hedging and repeated politeness, and sends the short version. Same request, same answer, fewer tokens. The counter next to the prompt box turns green with what you saved. Works on ChatGPT, Claude and Gemini. Free. No account. Everything runs in your browser. BEFORE "Could you please analyze this information and provide me with a detailed summary of the most important points, but please make sure that you focus specifically on the things that could potentially impact our marketing strategy going forward?" (~47 tokens) AFTER "Summarize the key points, focusing on what could impact our future marketing strategy." (~19 tokens) WHAT IT DOES • Runs on its own. Trims and sends when you press Send. The first time it pauses for a few seconds so you can see what it cut and choose "Keep on" or "Ask me first". • Never changes your meaning. Code, links, quoted text and e-mail addresses are left exactly as written. If a trim wouldn't be shorter, your prompt goes through untouched. • Keeps the words you choose. Like saying "please"? Click it once in the diff and TokenTrim never removes it again. • Shows every cut. Click the ✂ counter any time to see original vs. trimmed, word by word, and pick Light, Balanced or Max. • Suggests a faster model for simple tasks. A rewrite or a summary rarely needs a reasoning model, and those spend many times the tokens. When a thinking model is selected for a simple prompt, TokenTrim offers a one-click switch (Claude, Gemini) or opens the model picker (ChatGPT). Only when you click. • Shows what you save. The toolbar icon carries today's saved tokens; the popup has today, this month and all time. YOU STAY IN CONTROL • Turn automatic trimming off per site in one click, or switch to "Ask me first" and use the ✂ button when you want. • Want a moment to undo before each send? Set a 2 or 5 second window in Settings. Enter sends right away, Esc restores your words. PRIVACY • 100% local. Prompt text never leaves your browser and is never stored. • Only counts are saved: prompts trimmed and token totals, which you can reset any time. • No account, no tracking, no ads. No network requests at all. SUPPORTED SITES chatgpt.com, claude.ai and gemini.google.com. Token counts are estimates: every AI provider uses its own tokenizer, so exact numbers vary slightly. TokenTrim is an independent project and is not affiliated with OpenAI, Anthropic or Google.