Launch HN: VibeFlow (YC S25) – Web app generator with visual, editable workflows

[alepeak]

Hi HN! We’re Alessia and Elia, the founders of VibeFlow (https://vibeflow.ai). VibeFlow lets semi-technical people (i.e. people with some technical skill but who are not professional programmers) build full-stack web apps from natural language prompts, while making the underlying business logic clear and editable as a visual workflow. Demo video: https://www.youtube.com/watch?v=-CwWd3-b1JI.

The problem we’re trying to solve: today, people who want to build apps without coding often have to stitch together multiple tools, e.g. using Lovable for the frontend, n8n for workflows, and Supabase for the database. That creates data silos and leaves builders with fragile apps that break in production, don’t scale, and aren’t safe. We saw YouTube tutorials teaching people how to duct-tape these together just to get a functional app running. As engineers building no-code tools, we realized that people wanted the power of AI-generated UIs but also the ability to see and control their backend workflows and data.

Our solution is to generate the whole app at once, and represent it as a visual workflow. Users describe what they want in English (“I need a chat widget with an AI agent”) and VibeFlow generates both the interface and the logic. That logic shows up as a workflow graph they can edit visually or by giving new instructions.

We use Convex (https://www.convex.dev/) as backend. The generation of the backend code is fully deterministic, we map workflow graphs to code templates. This makes deployments predictable and avoids the hallucinated, black-box code you often get from AI.

Workflow representation: the logic is a directed graph where each node can be customized. We currently support CRUD operations and agent components. Any changes to the graph compiles directly back into code, so you always own the underlying logic.

Frontend: generated via AI and directly linked to workflow outputs, so it always stays in sync with the business logic. Changes to the frontend can be made through a chat interface.

Semi-technical builders can create maintainable apps (not opaque “magic”), and technical folks can still inspect the code and architecture. Compared to Bubble/Webflow, the interface is simpler; compared to Zapier, the workflows have an output code; and compared to AI coding assistants, you get an automatic backend plugged in with no black-box.

You can try it here: https://app.vibeflow.ai/. The demo video is https://youtu.be/-CwWd3-b1JI We'd love to hear from the HN community, whether you're a builder who's struggled with stitching tools together, a developer who's seen the pain points in no-code platforms, or someone curious about where AI-powered app generation is heading - we're eager for your thoughts!

[tchock23]

I want to like this and dig into it as someone who has recently used Lovable and Base44 (and been using Bubble for a while), but the YouTube ‘demo’ video is really weak.

The pace is too fast and you spend barely any time showing off your visual workflow feature, which according to your description is your differentiator.

I would strongly recommend using some of your YC money to have a professional recreate that demo and show off what makes you unique. Even if it goes longer than two minutes - if I’m interested I’ll keep watching.

I’ll still try it out because I’m a sucker for trying out new vibecoding tools, but you’re not doing yourself any favors with that video…

[alepeak]

Thanks a lot for the feedback. The video was meant as a very spontaneous ‘as it is’ showcase, but we’ll definitely make new demos that go deeper into the editor!

[echelon]

> recently used Lovable and Base44

Are you happy with either product? I tried them earlier in the year, and it was also really slow to make changes. I felt like they got stuck after a bit, too.

It's a neat concept, but I feel like they're expensive templates. I'd honestly prefer a template gallery with a smooth and fast editing UI.

[lysecret]

In general to me it makes a lot of sense to lean much more into "templates" (I'm sure lovable etc already do it, because it's also a nice way to save money). And it's much easier to at least guarantee some basic security when it comes to auth, payments, db setup etc. Of course you can shoot yourself in the foot right after that.

[alepeak]

Totally agree, security is a big point. It’s hard to trust LLMs on security, which is why we aim to make ‘white box’ backends

[lysecret]

Quite like the positioning of "this is the backend to your lovable ui", probably how chef (the vibe coding tool from the makers of convex) should have positioned it. (and kind of do).

[OldMatey]

This looks great! Can I export my end code / app and host it elsewhere easily? Where else would easily be able to host it?

[huevosabio]

They say they use Convex for the backend, which means you could in principle run it on your own account or go through the hoops of self hosting convex infra

[error404x]

I've been playing around with vibeflow for a while, it's impressive how fast you can go from a prompt to a working full stack app. The visual workflow editor is a game changer.

[alepeak]

Appreciate it! What did you build? What other nodes would be game-changers for you?

[Herobrine2084]

I think the evolution of vibe coding tool is definitely the editor. Having a black box with no way to maintain it is an absolute liability.

That's why I think app generators must be a good editor before being able to generate anything. It seems you went this way with the cool node interface.

I'm doing the same thing with https://luna-park.app, but for fullstack apps.

[Taig]

I'm seeing a huge union jack overlaying the page when I open it in Safari

[mguerville]

Bolt.new has a nice IDE and Roo.code is literally just a VS Code Plugin

[anakaine]

I've just tried my first prompt on your "try" setting, and what was generated wasn't anything like what I was expecting, to be honest. I asked for a workout tracking app with a good list of requirements (maybe it was too complex?), and the platform thoughts for quite a while. What I got was a template for a front end to connect to an AI agent for chats.

[replwoacause]

I didn’t have to scroll very far to find this kind of comment, which I was honestly expecting right from the get go. I feel like I’m always seeing this kind of feedback for these kinds of services.

[christoff12]

I like this. Any chance you'll be bringing similar tactility (is that a word) to the frontend? Granular changes to components via prompts leaves a lot to be desired.

[alepeak]

like visual edits to ui?

[bryanhogan]

Will have to try this later, the YT video looks promising. Found tools similar to this promising to create early mockups or other pre-prototypes when developing products.

[alepeak]

Thanks! Would love to hear what prototype ideas you have in mind

[dcsan]

I do like convex but do you support any other data stores?

[filipeisho]

Seeing the backend nodes generate feels like magic

[alepeak]

Glad to hear that. We want to make it as logical and white box as possible. Have you tried adding custom behavior after the first generation?

[deepdarkforest]

Congrats! Doesn't replit have an integrated database as well? Lovable has supabase, and I'm pretty sure Base44 as well, plus other agent integrations.

[alepeak]

Thanks! Yes, Replit has KV store and managed Postgres, Lovable uses Supabase (requires manual setup). Base44 doesn't have a manual setup but has a black box backend. In VibeFlow: - no manual setup required - low code backend editor n8n style - no black box anymore - everything you do in the backend is code that you own

It's not just about databases, think about all the users currently using n8n with Lovable separately, without even owning the full stack

[orliesaurus]

I tried this but kept getting errors. I asked it to build a TODO list that searches the internet to "augment" my todo list with advice

[alepeak]

what errors did you encounter?

[vinibrito]

Well I am/was building something that looks a lot like this, a shame I never applied to YC, wondering now if I should apply to other funds now so I can continue working on it, the prototype is ready so I have the main part figured out. A question, perhaps, could you give some tips to pitch this specifically, just for incubators, based on your experience?

[weird-eye-issue]

Why not focus your energy on selling it to real people instead of figuring out how to pitch it to incubators?

[Chaollapark]

Well all incubators ask the same thing (and search for the same profile). Just blast it to every incubator you find. Take some time write a nice YC application. There are tools like acceleratorfiller.xyz to send them to multiple accelerators. BTW (It's my company)

[johndevor]

[CONVEX M(events:insertEvents_ion)] [Request ID: bbc76cc0a8e100df] Server Error Called by client

[alepeak]

is any of the nodes in your editor marked as "incomplete"? there is probably a wrong call to the backend from your generated frontend. you can either ask the chat, or can help you debug in our discord https://discord.com/invite/Ctm2A2uEaq

[pzullo]

Why did you use convex as backend?

[alepeak]

Great question! We chose Convex for multiple reasons:

– We spin up isolated projects for each user. Convex handles this seamlessly with zero manual setup, while Supabase/Firebase have limitations and manual configuration needed – We abstract backend logic as visual nodes, so Convex's modularity makes it logical to find the right granularity for workflow representation. – Everything is reactive, so UIs and workflows stay in sync without bolting on listeners – Everything is end-to-end TypeScript with transactions by default, so generated code is predictable and maintainable

[PhatBrain]

nicely done. are you sideswiping n8n as well here? i like it.

[alepeak]

yes!

[fazkan]

congrats on the launch, lots of competition in this space. (leap.new, replit etc). Even convex has their own app-builder.

[alepeak]

thank you! There’s definitely a lot happening in this space, our focus is on making backends secure, robust, and understandable rather than just black-box codegen

[seanwessmith]

did you think about using Effect.ts as the backend? i'm interested in pros/cons there

[Babkock]

Awesome. Really, really awesome. One day no one will actually know how to write code, and no one will be able to fix all this shit.

[sirjaz]

Now make this build desktop apps. We don't need more web or mobile apps

[__natty__]

Don't get me wrong, I wish your start-up all the best, but this particular application seems so stereotypical by current standards. It's at least four buzzwords combined into one "idea". As someone who has never tried to apply, I wonder how difficult it was to get through Y Combinator's selection process.

[lagrange77]

That's the most 2025 startup name and idea i've come across so far.

[toddmorey]

I worry that almost all the 2025 startups I've seen are AI app builders. Where are the novel new applications? I get that codegen is currently one area where AI does well, but it also feels like we're struggling with other use cases.