I ran a benchmark that surprised me. Same AI. Same prompt. Two different approaches to building n8n workflows. Here's what happened: n8n MCP / Skills: ⏱ 15+ minutes to generate 🔴 Hit context limits mid-way ❌ Unconnected nodes, validation errors — unusable out of the box n8n-as-code: ⚡ ~3 minutes 🟢 Low token usage ✅ 100% valid, pushed live with no errors The reason isn't the AI — it's the architecture. MCP dynamically fetches node schemas on every call. Constant roundtrips, massive context pollution, and the AI still guesses wrong half the time. n8n-as-code maps the entire n8n node structure as TypeScript in your IDE. Your AI agent has perfect context locally — before it writes a single line. When to use which: → n8n MCP — Quick exploration, simple workflows, no IDE setup needed → n8n-as-code — Complex workflows, team projects, GitOps, production-ready automations Reliable automation isn't about more nodes. It's about tooling that doesn't hallucinate them. Has anyone else here made this switch? Would love to hear what workflows you're building with it.

Excited to share that out of all April submissions, my NPM Package Intelligence Agent was selected as a spotlight winner! 🚀 Built an AI-powered agent that analyzes any npm package and delivers data-driven recommendations using: — GitHub & npm APIs for real-time metrics — Firecrawl for web intelligence — Gemini for structured insights Helps you avoid risky dependencies before they hit production. Huge thanks to the n8n team for running such an incredible challenge — this community pushes builders to go deeper and build smarter. 🙌 Free template available — link in comments 👇

It's officially live on the n8n template hub! 🎉 For those who missed my earlier post — this workflow reads any PDF contract and flags every risky clause. Upload PDF → risk report in 5 minutes. No legal background needed. Grab it free 👇 https://n8n.io/workflows/15307

Most people open Claude Cowork and just... type. They're missing 80% of what it can do. Here are 5 ways I use it to actually get work done — not just generate text: → Auto-organize files without touching a folder → Send emails without opening my inbox → Summarize any doc in seconds → Schedule tasks that run on their own → Turn raw data into a shareable report Swipe through 👉 Which one are you trying first? Drop it below 👇

Most people use n8n to automate workflows. I used it as the backend of a live SaaS product. 💡 What I built Legal Contract Risk Analyser Upload a legal document → AI analyzes every clause → you get a full risk report in under 5 minutes. This is not a workflow sitting inside the n8n canvas. 👉 It’s a real web app anyone can open and use. ⚠️ Why I built this Most of us sign contracts we don’t fully understand. - 20 pages of legal text - 2–3 hours of reading - Still unsure what’s risky And the worst part? The riskiest clauses don’t look risky. Example: “The Client accepts full responsibility for all third-party claims” Looks harmless. It’s not. And keyword search? Completely useless here. 🧠 How it works This is where things get interesting. Instead of keeping everything inside n8n, I split it into: Frontend + Backend + AI pipeline 🔧 Under the hood: → n8n → runs the entire backend pipeline → Google Gemini → analyzes each clause, scores risk, rewrites it simply → Hybrid RAG (pgvector + BM25) → finds risky patterns using semantic + keyword search → Supabase → stores knowledge base + analysis results → HTML / CSS / JS → frontend UI → Netlify → makes it publicly accessible 🎯 What you get Every clause gets: ✅ Risk level (High / Medium / Low) ✅ What it actually means (no legal jargon) ✅ A safer alternative No more guessing. No more signing blind. 🧩 Who this is for - Freelancers reviewing contracts - Startups signing vendor agreements - Business owners who can’t afford hidden risks 🔥 The real lesson n8n is not just an automation tool. It can be your backend for real AI products. Same tool. Different thinking. Try it with any contract — link in comments 👇 If you’re interested in seeing the Product Requirement Document (PRD) and how I built this end-to-end,drop your email in the comments — I’ll share it.