Prompt Optimizer

Hey Product Hunt! 👋 I'm Dwelvin. Creator of the "Prompt Optimizer". I’m super excited (and a little nervous 😅) to share Prompt Optimizer — a tool born from late-night experiments, API rate-limit warnings, and way too many “why is this prompt not working?” moments. 🎯 The Problem About a year ago, I was terrible at prompting. Like… genuinely bad. (Still not the best, but much better). I’d spend hours trying to get Claude or GPT-4 to give me coherent results. I’d hit rate limits, burn credits, and still end up rewriting the same prompts. As a solo dev without formal writing chops, it was frustrating and expensive. Every time I thought I’d cracked the code, a new model came out — GPT-4, Claude, Gemini — each with its own “personality.” What worked for one completely failed on another. Image prompts? Whole different language. I realized I was spending more time fighting the prompt than building the actual product. 💡 The “Aha” Moment One day it hit me: I couldn’t be the only one. Developers, writers, designers — everyone was just guessing their way through AI tools. We needed something that could think about prompting for us — something that learned best practices and turned trial-and-error into repeatable wins. So I built Prompt Optimizer to do exactly that: stop guessing, start engineering. 🛠️ What I Built 🌐 Two Modes, One Ecosystem 1️⃣ Local Engine (Privacy-First) Works 100% offline — no dependencies, no cloud calls Runs 127 optimization rules locally Free tier: 5 optimizations/day, no signup Great for sensitive projects or learning the ropes 2️⃣ Cloud Service (AI-Powered) Detects your intent (code, image, or LLM prompt) automatically Builds SOPs and reusable “prompt templates” on the fly Auto-saves high-confidence results (>70%) Bring your own OpenRouter key and tap into 100+ models 🔄 Universal Integration Works out of the box with Claude Desktop, Cursor, Windsurf, VS Code, Zed, and 10+ other tools — no copy-paste, no switching windows. 📚 What I Learned Along the Way Building this taught me so much more than I expected: ✅ How real users actually talk to LLMs (spoiler: not how you think) ✅ Why context > cleverness in prompt design ✅ The beauty of technical writing — clear beats fancy ✅ Cross-platform dev pain (hi, macOS ARM 😅) The biggest takeaway? Prompt engineering isn’t an art — it’s a system. And systems can be automated. 🎁 What’s Inside Free Tier (No Signup) Install the local engine via npm 5 free optimizations/day Full 127-rule optimizer MCP-$19.99 (one time payment) own forever MCP-Local Template Saving with Metadata. MCP-Template retrieval, Stats, Use a base. Works offline forever Paid Plans ($2.99–$69.99/mo) 5,000–75,000 optimizations/month Cloud intelligence & analytics Team libraries and shared templates SOP + skill package generation Context Engineering App automates Standard Operation Procedure document creation. Create custom "Skill" with a prompt, url or context from uploaded documents (up to 50MB). 🚀 Try It Out Quick start (2 minutes): npm install -g mcp-prompt-optimizer-local 📘 Full docs: promptoptimizer-blog.vercel.app Thanks for checking this out, Product Hunt! 🙏 If you’ve ever felt frustrated with prompting, this one’s for you! Try it out with promo code PHD48 for 48 hours, complete access, to all tools, no charge! Promo expires 7 days after redemption. Would love your feedback — what’s your biggest “AI prompt pain point” lately?