Less Tokens — Prompt Compressor for ChatGPT, Claude, Gemini, Copilot

About

Requires the free less-tokens backend running on your own machine: `pip install less-tokens`, then run `less-tokens-serve`. All compression happens locally — your prompts never leave your computer.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Less Tokens is a prompt compressor for AI chat. It trims the filler, stopwords, and grammatical scaffolding that models quietly ignore — typically cutting 30–40% of the tokens in a prompt while keeping the answer essentially the same. Fewer tokens means lower cost and faster responses, especially if you're working against paid API limits or long context windows.

A "less tokens" button appears right next to the chat box on the sites you already use. Click it, compose or grab your prompt, compress, and insert the shorter version back into the box — ready for you to review and send.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
WORKS with known LLMs and AI applications

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
WHAT YOU CAN DO

- One-click compress — grab whatever is in the chat box and shrink it in place.
- 11 compression techniques — toggle exactly which passes run (filler phrases, stopwords, contractions, abbreviations, lemmatization, and more). Negations and question words are never dropped, so your intent stays intact.
- Zone-aware compression — split a prompt into "free" (compress fully), "careful" (safe passes only, for rules), and "protected" (left untouched, for JSON schemas or examples), each color-coded.
- Attach files & images — pull clean text out of a PDF or Word doc, or OCR an image, then include only the words instead of the whole file.
- See what you saved — every result shows the token reduction before and after.
- Insert into chat — drop the compressed prompt straight back into the site's input box.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
PRIVATE BY DESIGN

The compression engine is the open-source `less-tokens` Python package, and it runs entirely on YOUR machine. The extension sends your prompt text only to http://localhost — the backend you run yourself — and gets the compressed result back.

Your prompts, documents, and images are never sent to us, never stored, and never used for training. The only data tied to your account is your name, email, phone (optional), and an encrypted password — used solely so you can sign in as a less-tokens community member.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
SETUP (one time)

1. Install the package:  pip install less-tokens
2. Start the local backend:  less-tokens-serve   (runs on http://localhost:8000)
3. Sign in to the extension with your free less-tokens account.

Once the backend is running, the status indicator turns green and you're ready to compress. If it's not running, the extension simply shows "backend offline."

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
OPEN SOURCE

less-tokens is MIT-licensed and fully auditable — the whole compression pipeline is deterministic, training-free, and runs on CPU.

- Package: https://pypi.org/project/less-tokens/
- Source: https://github.com/shaminchokshi/less-tokens
- Website: https://www.lesstokens.org/