PromptWise by Nirupam Chakrabarti

PromptWise helps you get the same results from AI assistants while spending fewer tokens.
As you type in ChatGPT, Claude, or Gemini, PromptWise reads your draft and suggests a tighter version — removing filler and politeness, simplifying verbose phrasing, and dropping context the model already has. You see the rewrite and the exact token saving, and apply it with one click. Nothing is ever sent automatically.
It also watches your conversation and nudges you when:

• the thread has grown long, repetitive, or off-topic — offering to summarize and start a fresh chat, and

• you're using an expensive model for a simple task — suggesting a cheaper one.
Privacy first: the engine runs entirely in your browser and makes no network calls. Your prompts never leave your device, and PromptWise collects and transmits no data.
Optional local AI: if you run Ollama, or enable in-browser WebLLM, PromptWise can produce higher-quality rewrites and real conversation summaries — still 100% on your own machine, nothing sent to any server.
Works on chatgpt.com, claude.ai, and gemini.google.com.
Free and open source (MIT): https://github.com/Nirupam014/promptwise