Join Us! | Try Unstuck | Other Articles
This Wednesday, I sat down with an idea and Claude (obviously) and 5 hours later I had an app live on the internet. With user authentication, email integration, scheduled backend jobs, security checks, and ai integration without having to leave my Claude chat window.
Those of you who know me, know that I am a professional vibe coder.
I build everything using either Claude code or Cursor or sometimes both. And I have never really liked Lovable, Replit, or Base44 cause every time I use it, I feel like I am building something on their terms.
Their platform. Their credits. Their partners.
Floot was the first one that didn't ask me to move.
But this is NOT a Floot is amazing post. This is my honest review including things that went wrong.
Because founders today are bombarded with AI tools to use. So you need someone in your corner who is testing these tools and telling you whether they are worth your precious time.
Quick disclosure before we go ahead:
This is NOT a paid post. I did get access to Floot’s highest tier to try out the product and my paid subscribers are getting 10% off on Floot. But everything I say here is my honest opinion.
What Is Floot?
Floot is an AI app builder that gives you a real product without making you integrate with any external services. Meaning, you get a database, user authentication, hosting, email integration, ai integration, and scheduled job on a single platform.
So you basically use Floot with your AI subscription to build your MVP. No pesky integrations to look at. No Github, Supabase, Vercel etc etc.
Your AI is the brain: It makes the decisions, spots the issues, and argues with you
Floot is the body: It holds the database, sign in, email, hosting, jobs and gives your AI the tools/guidelines to build your app
But this also means that the quality of the decisions relies on the model you choose. So built on two different models, Floot can technically give different outcomes.
How Did I Use It?
Before I build anything, I brainstorm. This is a standard practice I have and the most important reason why I can vibe code basically anything very quickly.
Once I was set on an idea, I asked this brainstorm chat window to create 4 documents:
A Design Document: This dictates the UI of the product. The colour scheme, typography, spacing, formats etc etc.
A Product Document: A brief on what I am building, who is it for, what will it do, what is the user flow, what screens will exist, how will everything interact, and what I absolutely do not want to build.
A Data Document: A brief on every piece of data the product needs, how it interacts with each other, and where is it used.
A Build Order: A stepwise guide on what needs to be built, what will that step entail, what checks need to happen, and when.
Now normally I have a few more documents that I create but for Floot I could do with just these four.
And I kept the chat I used to create these documents to give instructions to Floot.
What I Liked
You can trust Floot with your database and security. Without me giving any guidance on the database, Floot gave me a verified user sign in, stored encrypted passwords, enabled account lockouts, and established sessions tracking. So the issue that a non-technical builder might have with supabase where one user can see other user’s data, will not happen with Floot. And I didn’t have to do anything for it.
No pesky integrations to handle. Usually I would have to connect Vercel, Github, Google Auth, Supabase, Resend, Anthropic etc etc to make my backend work. But Floot handles everything. So a non-technical founder can actually put a proper app in front of their first users.
It follows the scope at all times. My first message started with the four documents that I mentioned. And any time any of my instructions contradicted those rules, Floot immediately notified me. Most of my ai builds take random detours because the scope quietly expands until you can’t keep a track of things. But this one did not.
It follows a design document religiously. Similar to the scope logic, most of my ai projects can tend to forget the design rules that I had set early on. But Floot, while giving me the room to be creative with design, made sure that whatever rules I have set are followed. But it was mainly because I did create a design system to begin with.
It triple, quadruple checks everything. Every time I asked Floot to do anything, it checked it against the rules I had set and gave me a very detailed report. I asked it to check 3 specific edge cases, it checked for 9. So for a non-technical builder who does not know what to check for, giving a simple prompt like “check for edge cases” will result in a thorough review.
What I Disliked
If it is not in the rules you mentioned, it can sometimes not ask you before making a decision. For instance, I did not have anything about selecting an AI Model. So Floot chose the best one which would have made my user wait for 4 minutes to get a response. Another example, my product has people talk about their current bottleneck and I did not mention how I want that data handled. Floot, stored all the transcripts indefinitely and sent the entire data to both Open AI and Gemini to process. So, you have to be a bit vigilant about the things you did not dictate. After a few builds, if I asked Floot something like, “what did you decide for me that I did not ask for”, it started helping me see all these gaps.
It is not very intuitive to understand which version of your product is live. Both the sandbox and the live product get data and information from the same place. So if your live version is not the latest one, you can see a broken product. The way out is to make sure you republish any changes as and when you make them.
A non-technical builder still needs to know the legalities of distributing their product. And this is the case with every AI app builder. No one will tell you the legal compliance, detect missing pages, or find data leaks if you don’t explicitly ask it to. So, the burden of this still relies on your shoulder.
I would say Floot is a tool that will build exactly what you ask for.
TLDR
Verdict: Use It
Best For: A non-technical founder who wants a real MVP live in a few hours and doesn’t want to deal with any additional service providers
Why It Wins: Everything is in one place and it uses your existing AI subscription. So you don’t have to manage multiple platforms or multiple subscriptions or do complex credit math. Floot uses build actions to complete a task. And one action is pretty generous. It can get a lot done, depending on how you prompt.
What To Know: It build what YOU ask for and does not have guardrails for unexpected behaviors. So your prompts have to be good. Your model selection also matters. You can download your code and your database at the end of everything, but you do have to check and ensure that nothing in the codebase points to anything in Floot. So it is maybe a few hours of work if you want to migrate.
Skip It If:
You can already code (duh!). You are much better off with Claude/Cursor/Codex.
All you need is a landing page. The UI is pretty basic and you might generate a much better result in Claude code or Codex.
P.S. I did not try integrating external libraries for any UI related tasks. So next week I might do that and write a followup here.
What Did I Build?
This is something I am EXTREMELY EXCITED for!
I have built a platform for early-stage solo founders who are stuck on a problem.
Share your problem with Unstuck and the platform matches you to another founder who is stuck on the same or similar issue.
The platform is LIVE. So if you are stuck on a problem right now, you will find a match within 24-48 hours.
Try it here!
What Is Below The Paywall?
A comparison between Floot and all other available AI builders
A guide on how to write better prompts to use with Floot and the exact prompts I used
A 10% discount for the first month of Floot
