The quality-control platform for AI prompts.
Your prompt, scored.
Then rewritten properly.
Paste any prompt and get a quality score across seven categories, the exact issues holding it back, and an optimised rewrite built for ChatGPT, Claude, Gemini or Grok. Use Local Rules free in your browser, or bring your own provider key for AI-generated analysis.
Local Rules: free, no sign-up, unlimited · history stays on your device
Popular categories
How it works
Paste your prompt
Pick your target model, the task type, and how deep the optimisation should go.
Get the report
A 0–100 score across clarity, context, specificity, constraints, structure, format and reasoning — plus the specific issues found.
Copy the rewrite
An optimised version written for your chosen model, with a before-and-after comparison so you can see what changed and why.
Three depths of optimisation
⚡ Quick
Light-touch fixes in seconds. Tightens wording and fills the obvious gaps.
🧠 Professional
Built for real work. Adds context, a clear role, structure, formatting instructions and constraints.
🚀 Expert
A full engineering template: delimiters, placeholders, validation steps, output schema and edge-case handling.
Questions
Is it free?
Local Rules is free, in your browser, with no sign-up and no daily limit. If you want AI-generated analysis instead of the deterministic checks, you can bring your own OpenRouter, OpenAI, Anthropic or Gemini API key — that usage is billed by your provider, not by us.
Which AI models does it optimise for?
ChatGPT, Claude, Gemini and Grok, plus a Universal mode that works well across all of them.
Do you store my prompts?
Local Rules never sends your prompt anywhere — it's analysed entirely in your browser. Your history is stored on your own device. If you choose a direct provider key instead, your prompt is sent straight to that provider's API from your browser tab.
How is this better than asking ChatGPT to improve my prompt?
You get a structured report — a score across seven categories, the specific issues found, and a rewrite built for the model you're targeting — rather than a single unexplained rewrite.
See what your prompt scores
It takes about fifteen seconds, and the first thing most people learn is which category they always skip.
Open the optimiser