# BeforePrompt > BeforePrompt is building the intelligence layer before AI acts. Today it runs inside ChatGPT, Claude, Gemini, and Microsoft Copilot, where it checks an instruction locally, identifies missing context or risk, and offers deliberate actions before the user sends it. Key facts: - Works with ChatGPT, Claude, Gemini, and Microsoft Copilot via a Chrome extension. - The AI Intent Gap is the difference between what a person wants accomplished and what an AI system actually understands, receives, and executes. - BeforePrompt's current browser product addresses part of that gap through intent checks, context guidance, sensitive-data warnings, and user-requested actions. - The quality check runs locally in the browser; a cloud model is only used when the user clicks a rewrite action (Improve, Shorten, Structure, Clarify). - Its supporting library includes 1,000+ approved prompts across 28 work domains. - Fifty complete prompts across 26 domains are public and free to copy; each has an expert role, named inputs, a defined output, and practical limitations. - Private by default: drafts are not sent anywhere to be scored; raw prompt text is stored only with explicit opt-in. - Browser support is available now. Broader desktop, IDE, CLI, local-model, API, and agent integrations are platform direction, not presented as currently shipped. - Company: BeforePrompt. Site: https://beforeprompt.com. Company enquiries: info@beforeprompt.com. General feedback: hello@beforeprompt.com. Support and privacy requests: support@beforeprompt.com. ## Pages - [Home](https://beforeprompt.com/): what BeforePrompt does and how the flow works. - [Chrome extension](https://beforeprompt.com/extension): supported AI sites, features, and the privacy model. - [Free prompt library](https://beforeprompt.com/prompt-library): 50 public prompts across all 26 domains, plus an introduction to the signed-in catalog of 1,000+ prompts. - [Capability status](https://beforeprompt.com/capabilities): the source of truth for available, testing, exploring, and long-term capabilities. - [Resources](https://beforeprompt.com/resources): the AI Intent Gap framework, privacy architecture, and benchmark methodology. - [About](https://beforeprompt.com/about): why BeforePrompt exists, its principles, and its independent operating approach. - [Support](https://beforeprompt.com/support): contact form for product, account, billing, privacy, and team-workspace help. - [Privacy](https://beforeprompt.com/privacy): how prompt data is handled. ## Blog Practical, honest guides on writing better prompts for ChatGPT, Claude, Gemini, and Copilot. - [How to write better ChatGPT prompts](https://beforeprompt.com/blog/how-to-write-better-chatgpt-prompts): four habits — lead with the request, name the output format, add the context the model can't guess, show an example. - [Why your first prompt is usually your worst](https://beforeprompt.com/blog/why-your-first-prompt-is-your-worst): the re-run tax, and a ten-second pre-send check that replaces several follow-up messages. - [Prompt checker vs. asking the AI to improve your prompt](https://beforeprompt.com/blog/prompt-checker-vs-asking-ai-to-improve-prompt): an AI rewrite polishes wording after you send; a pre-send checker flags missing context and pasted secrets before you send. - [Is it safe to paste sensitive data into ChatGPT?](https://beforeprompt.com/blog/is-it-safe-to-paste-sensitive-data-into-chatgpt): the real risk is you can't un-send it; what never to paste (keys, PII, NDA material), why turning off training isn't enough, and catching secrets locally before sending. - [How to write better Claude prompts](https://beforeprompt.com/blog/how-to-write-better-claude-prompts): Claude takes instructions literally — use XML tags for long prompts, put context before the task, explain why, and show examples. - [How to write AI image generation prompts](https://beforeprompt.com/blog/how-to-write-ai-image-generation-prompts): stack subject, setting, style, lighting, camera, and aspect ratio; 30–50 words beats three; Midjourney wants keywords, DALL-E wants sentences. - [How to write AI video generation prompts](https://beforeprompt.com/blog/how-to-write-ai-video-generation-prompts): describe action, camera movement, continuity, duration, and style; keep camera moves slow (fast still warps); Sora/Veo/Runway each want a different order. - [How to write better Gemini prompts](https://beforeprompt.com/blog/how-to-write-better-gemini-prompts): Gemini rewards short, direct prompts with no filler; use Persona-Task-Context-Format and refine over a few rounds. - [Why the same prompt gives different results in ChatGPT, Claude & Gemini](https://beforeprompt.com/blog/same-prompt-different-results-chatgpt-claude-gemini): each model is trained with different instincts; how to adjust one prompt for each while keeping outcome/context/format constant.