Show HN: My colleague said my prompts were unreadable. I built a prompt builder
flompt.devLast week I started using Claude Code. My colleague, who has been prompting AI models for months, looked at what I was sending and said he had no idea what I was asking for.
If an experienced user couldn't parse it, the model definitely wasn't getting the best version of it either.
So I built flompt. The idea is simple: instead of writing a prompt as a wall of text, you decompose it into typed visual blocks (role, context, objective, constraints, examples, output format), arrange them, and compile to a structured format optimized for the model you're targeting. Claude gets XML, ChatGPT and Gemini get Markdown, each formatted the way the model actually processes it best.
It also ships as a Claude Code MCP so you can build and inject structured prompts directly from your editor, and as a Chrome extension that puts it as a sidebar inside ChatGPT, Claude and Gemini.
A few things I cared about: - Assembly is 100% local, nothing leaves your browser during compile - No account required - You can paste a rough idea and let the AI decompose it into blocks automatically
Try it: https://flompt.dev/app Chrome extension: https://chromewebstore.google.com/detail/mbobfapnkflkbcflmed...
https://flompt.dev Hey HN, builder here. Happy to answer questions. The prompt structure I landed on after testing: role → context → objective → constraints → examples → output_format. That ordering follows Anthropic's guidance and consistently outperforms reordered variants in my tests. The MCP server is probably the most developer-useful part — decompose_prompt() and compile_prompt() are callable from Claude Code agents, so you can use flompt to structure prompts dynamically in agentic pipelines rather than just for one-off prompting. The visual canvas (React Flow) was a deliberate choice over a form-based UI. Seeing prompt blocks as connected nodes makes it obvious when you're missing something (e.g., no examples block when you have complex output requirements). Free, no account, open source: https://flompt.dev