Most disappointing AI outputs trace back to a vague prompt. “Write a blog post about our feature” leaves the model guessing at audience, length, tone, and structure. This tool takes your rough prompt and rewrites it into a clear, structured instruction using your own OpenAI or Anthropic key — and tells you exactly what it changed.
How it works
Pick a provider and model, paste your API key, and choose which model the improved prompt is meant for. Paste your current prompt and, optionally, your goals for the rewrite. The tool sends a meta-prompt that instructs the model to restructure your prompt into Role, Task, Context, Constraints, and Output-format blocks, replace vague language with specific instructions, and add explicit success criteria. It makes one direct request to the provider and returns the improved prompt in a fenced block plus a short changelog of what it altered and why.
For Anthropic, the request includes the official direct-browser-access header so it works straight from the page.
Why structure beats length
A long prompt is not the same as a clear one. The blocks the improver enforces map to the questions a model silently asks: who am I, what is the job, what context do I have, what must I avoid, and what should the answer look like. Answering those explicitly removes the guesswork that produces generic output. The changelog is the most useful part for learning — read it to see which weaknesses the model spotted, and you will start writing tighter prompts yourself.
Tips and limits
Run an important prompt through once, read the changelog, then tweak the result rather than accepting it blindly — you know your context better than the model does. Keep any concrete facts, names, and examples from your original prompt; if the rewrite drops them, paste them back into the Context block. And remember the improver optimises clarity and structure, not factual accuracy: it cannot know whether your underlying request is the right thing to ask in the first place.