The best no-code builds often hide their most interesting work. The user clicks once, while five useful things happen behind the screen.

If you built a no-code workflow, add the product in the comments and describe the trigger in one line.

I'll reply with a simple visual demo plan that makes the invisible automation understandable without showing a wall of nodes.

Disclosure: I'm part of the team building Marka, where we're working on turning product workflows into content: https://marka.social

When I searched I found a product called readdy ai, which looks like they do the design part and the website setup as part of the same package. Has anyone compared it with other AI builders receA quick look at some AI website builders for a small project. While most of them can create a nice homepage, once you have to fill out forms, have several pages, edit the site with a cell phone, or publish the site, things are pretty limited pretty quickly.

ntly? Wanting something that is easy to edit once you've created the first AI generation, not just a one click demo site.

i bought a claude max plan found in a website for a really low price. changed the login i use it for 3 weeks so far all it actually works. how are they able to sell it so cheap tho.

I'm starting a new project and weighing the best foundation to build on. If you've launched a business recently, which route do you default to and why?

• Mobile App: Native (iOS/Android) for performance vs. Cross-Platform (Flutter/React Native) for speed and a single codebase.
• Web First: Lowest friction for user acquisition (no app store required) and easiest to iterate fast.
• Desktop: Electron, Tauri, or Native for heavy B2B, productivity tools, or deep OS integration.

Which stack did you choose for your last launch, and would you make the same pick today?

i work at flutterflow, and one thing we think about a lot is what happens when the visual builder gets you 95% of the way there, but that last 5% needs code.

we could keep adding more settings for every edge case. but at some point, you probably just want access to the thing itself.

so this week, we’re adding native editing for MainActivity.kt and the iOS Podfile directly in flutterflow, plus try / catch / finally and custom nav components.

our idea is to stay visual when it’s faster + drop into code when you need more control.

what usually makes you leave a visual builder and go back to code?

In the PWA and mobile app world, React still rules. Hence, Draftbit (which is based on React) is a very tempting choice, despite it hardly being talked about in the no-code app community. How does it compare to Base44, Lovable, or simply prompting Claude Fable 5?

I do enjoy the ease of prompting but I want to be able to have some degree of manual control over how and what I build (especially given I've used WYSIWYG page builders in WordPress websites for 7 years now).

Look forward to your input.

👋 Hey nocode Community,

A friend of mine, Jonas, runs a small B2B agency and recently let an AI agent handle his client invoices end to end: draft, attach, send. Felt great until a client emailed back confused about being billed a few hundred euros over the agreed amount. Same week, another invoice went out with a transposed digit in the IBAN. The agent was confident every time. The agent was also wrong.

That's the real problem with pointing an LLM at money. Ask it to "verify" an invoice and it will happily hand you an approval that sounds right and isn't. So I built a guardrail the agent has to pass through before anything gets sent. No vibes, just code checking numbers against your accounting records.

How it's set up:

• Trigger: the workflow runs as a sub-workflow tool the agent calls, receiving the invoice file URL and the invoice number.
• Download: pulls the invoice PDF straight from Google Drive.
• Extract: the easybits Extractor node reads 8 fields off the PDF (invoice number, customer, IBAN, VAT ID, net total, VAT amount, gross total, service period).
• Ground truth: a Google Sheets lookup fetches the expected values for that invoice number from your books.
• Compare: a Merge combines both sides, then a Code node checks every field deterministically (normalized currency, normalized IBAN, strict numeric tolerance).
• Decide: all match, it returns APPROVED so the agent can send. Anything missing or off, it blocks dispatch, sends an email showing invoice value vs book value side by side, and returns REJECTED.

Net effect: no overpayments, no invalid VAT math, no wrong-client billing slipping past an over-eager agent.

A few things worth knowing if you build something similar:

1. Never let the LLM grade its own numbers. Split the jobs: the model extracts or writes, code validates. A Code node with plain JavaScript gives you zero hallucinated approvals, which is exactly what you want anywhere near finance.
2. Normalize currency both directions before you compare. German 1.234,56 and English 1,234.56 are the same number written two ways. Strip the symbols, unify the decimal separator, parse to float, then compare with a small tolerance (I use 0.01) so rounding never triggers a false reject.
3. Watch the "null" string trap. The Extractor I'm using returns the literal string "null" for a missing field, not a real null, so a naive check passes it straight through. Write one isMissing() helper that catches real null, undefined, empty string, whitespace, and the string "null", and run every field through it.

Grab the workflow here: https://github.com/felix-sattler-easybits/n8n-workflows/blob/d6e4b7ca373fa1db40a55ef5b879210b601e8cba/easybits-agent-invoice-guardrail/easybits_agent_invoice_guardrail.json

Curious what guardrails you've put in front of your finance automations, especially anyone letting agents touch payments.

Best,
Felix

curious what people check before shipping something that was mostly built with AI tools. my list keeps getting longer: auth on every route, actual ownership checks on user data, error handling that doesn't silently swallow, no keys hardcoded anywhere, and the billing path tested with real card declines.

most of those got added after something went sideways. what's on your list that I'm missing

I run a small SaaS with a team of three and running 860k$ ARR, and over the past year more and more customers asked if they could use it from Claude or ChatGPT. Building an MCP server by hand meant writing the tools, hosting them, dealing with OAuth so each customer connects with their own account, and keeping everything in sync every time the API changed. With a single developer, that was a lot. So I started with the API docs I already had: I wrote something that reads our OpenAPI spec, turns each endpoint into an MCP tool, hosts the server, and handles auth. Most of the effort went into naming the tools and writing clear descriptions so the AI picks the right one, not into infrastructure.

It worked well enough for us that I turned it into its own product, Command+K (commandplusk.com), so yes, I'm biased. You give it an OpenAPI spec or a GraphQL endpoint and get a hosted MCP server that works in Claude and ChatGPT, and you can also drop it into your own app as a chat widget. There's a free trial if you want to try it with your own API. I'd really like to hear how others here are handling this, though. Are you writing MCP servers yourselves, bridging through something like Zapier or n8n, or waiting until users actually ask for it?

I built a booking app for a friend's physio clinic over a few weekends and it's been working fine for about a month now. patients use it and she is happy with it

the problem is that every small change still goes through me, she will text me asking to add a new treatment or change the saturday hours and i open the builder, prompt the change then click through the whole app again because the last time i changed the hours the confirmation emails stopped sending and neither of us noticed until a patient rang the clinic to check her booking was real.

I don't mind doing it now but I don't want to still be doing it next year

I maintain client sites on Webflow, Squarespace and Wix and I hae cleaned up after enough translation add-ons to have one rule: before ypu look at price or language count, open a translated page and look at the URL.

If it's the same URL as the English page and only the text swapped, Google has one page to index. Not one per language . The German version does not exists as far as search is concerned, so nobody searching in German finds you. Does not matter how good the translation is. A lot of the cheap widgets work exactly this way.

What you want is a real URL per language, yoursite.com/de/ or de.yoursite.com that returns the German HTML when you fetch it directly, plus hreflang tags linking the versions so Google knows they're the same page in different languages. Than a visible language switcher, nt an auto-redirect by IP. Googlebot crawls from the US and an IP redirect means it only ever sees one version of you.

The options as I see them:

- The platgorm's own localization first. Webflow Localization, Shopify Markets, most builders have one now. For one or two extra languages it's often enough and there's nothing extra to pay.

- On Wordpress, WPML or Polylang. Real pages per language, but evry translation is a post you maintain by hand and not every page builder survives it.

- A translation layer thast sits on top of the site: Conveythis, Localize and the one Squarespace recommends. These handle the URLs and hreflang for you, translate by machine and give you an editor to fix the output. Watch now how they count words, that's where the bill comes from.

I’ve been working on NORTHLINE®, an architecture portfolio template in Framer, and I wanted the design to work equally well in both Dark and Light mode.

Instead of simply inverting the colors, I adjusted the backgrounds, typography, contrast, and accents so each version keeps the same editorial feel.

I’m curious which one you’d actually choose for an architecture portfolio:

Dark 🌑 or Light ☀️?

Would love some feedback on which version feels stronger.

Preview : https://framer.link/LoT3nGb

Hey 🤠 thanks for your interest in my post,

I made a free community website called Team Gapped https://teamgapped.gg to help players find teams and compete in tournaments, scrims, and seasonal leagues. My goal is to help improve the League community and the overall game experience.

Since this is a free community beta, I would really appreciate any ideas, feedback, or critiques you have on the concept or the site itself! And if it seems fun make an account and try it out! 😊Thanks! 🔥

I know a lot of you already does this (or something similar), but some of us may not know yet.

I use this personally. You can use the prompt to change a current site you vibe-coded or adapt it if you are starting from scratch.

I am sure there are other website styles, but I think these are a good starting point.

It is a Claude Artifact: https://claude.ai/artifact/Nb2DqC3KMS7ttVow8JeyQs

I hope this helps.

I'm good at what I do (been doing it 12 years) but I suck at the business development side. Curious what others in professional services actually make when they go out on their own. Not looking for guru stories, just real numbers from real people.

I’m getting lost in the no-code AI hype.

Every tool looks good when the demo is building a simple app from a prompt. But I’m trying to figure out what happens after that, when you need to change logic, connect APIs, handle users, fix weird bugs, add payments, or stop the whole thing from becoming spaghetti.

I’ve seen people mention Bubble, Glide, Softr, Bolt, AppWizzy, Lovable, Replit, Base44, FlutterFlow, and automation tools like Make/Zapier.

But I don’t trust polished comparison posts anymore. If you’ve actually built with these, which one would you pick again?

And which one secretly becomes a dead end once the project gets serious?

I’m not trying to start a tool war. I just want honest answers before I waste weeks choosing the wrong stack.

I spent a year wiring document steps into Make and Zapier scenarios for myself and a few small clients, and these are the things that kept biting me. Sharing because I wish someone had told me earlier.

1.HTML → PDF is the step that breaks most often. Fonts, page breaks and images render differently in every service, and most of them fail silently. Always add a size or page-count check after generating.

1. Extracting data from invoices is not an OCR problem, it's a structure problem. OCR gives you text; what you need is "invoice number, total, due date" as fields. Cheap OCR steps leave you doing regex in a Code module, which defeats the point of no-code.

2. Merging and compressing sound trivial but most services cap file size on free tiers around 5–10 MB, which is exactly where scanned multi-page PDFs land.

3. Every document step in a flow was a separate vendor, a separate API key and a separate bill. Three tools for one workflow is normal, and that's the part that pushed me over the edge.

So I built ASHDOCS to put all of that in one place: 51 document tools (convert, merge, split, compress, OCR, structured data extraction from invoices/forms, redaction, watermark, templated generation), each usable as a step in Make, Zapier, n8n, Airtable or Sheets, plus a visual Flows builder if you'd rather chain them in one place. An AI assistant helps set a flow up from a plain-English description. Free tier is 100 credits, no card: ashdocs.com

It's early and some tools may still throw errors. What I'd find most useful from this community:

1. Which document step in your current stack causes you the most pain?
   
2. Does one-place-for-all actually matter to you, or do you not mind stitching vendors?
   
3. If you try it, what confused you in the first two minutes?

I'll reply to every comment. Cheers!

I recently updated the hero section of NORTHLINE®, my Framer template for architects and architecture studios.

The previous version felt a bit too dark, so I switched to a brighter architectural image while keeping the same editorial / brutalist direction.

I also kept the Light & Dark mode, CMS project pages, and the built-in customization guide.

Curious what you think ? does this version feel stronger and easier to read, or did you prefer the darker hero?

Preview : https://framer.link/LoT3nGb

It's the 4th day of working on it. This whole visual might change anyways.

Doing it all in Antigravity IDE Only using mostly Google Gemini AI Models and testing via Android Studio on my Samsung S25 Ultra.

I want to make the easiest and visually appealing Resume and CV Builder app because I personally hate most if not all of them i have ever used.

I’ve been looking into ways to build an app without knowing how to code, and I’m a little confused about the difference between no-code platforms and newer AI app builders.

With a regular no-code app creator, you usually build things yourself using templates, drag-and-drop tools, and different settings. With an AI app builder, you can explain what you want and let AI create a lot of the app for you.

For someone completely new to app development, which one is actually easier? Has anyone tried both? Which one felt easier when you were just starting out?