I checked Base44 out to see what the hype all about. Spoiler: It Won’t Be All Rosy for Non-Tech Users.
A "honest" review of one of the fastest growing no-code app builder platform out there
This article is also published at my Medium
Alright, I admit the title is a bit click-bait-y. Full disclosure: I am technical BA and also have been in the tech scene for quite awhile now. I myself have been digging in the vibe-coding / assisted coding agent for the last 4 months and have had my hands on pretty much of popular buzz-worthy tools and agentic coding models out there. But I wrote this review article wearing the hat of those that still new to the “build-an-app-just-with-your-words” sales pitch and hope this will be useful for you. Also this post is not sponsored by Base44 nor its competitors (obviously, look at the title) and I don’t have any affiliation with the business
Intro
Once in a while I want to check out a new tools. I heard about Base44 on this Lenny’s podcast. For those who don’t know, in short Base44 is a online AI application building platform that allow user to quickly stand up a production-ready application within just a few hours of prompting. I personally resonate and admire a lot with the founder — Maor Shlomo , as I also love to build application to solve my pain points and hope one day one of that application can be a $$$ unicorn (may be that day still far away 😅). The hype is that Base44 is perceived really well by large builder community and that Maor himself make million selling the platform just 6 months after scaling it to impressive 400,000+ recurring users.
See / hear the full pod cast here
It happens so that a week ago, I have just had this new idea for a AI micro-Saas tool to help with context management so I thought “Why not trying out Base44 to see how quickly it can turn that idea into a functioning app — as they advertised”. “Oh well, what’s the worst that could happen?” I told myself, before diving right in.
First looks
Base44 has typical SaaS tier subscription structure allow user with different scaling needs to use their AI-app building capability. I would highly recommend to suggest you to go for the Builder pack because with lower tier like Free / Starter, you are going to run out of prompt token really quick (and there is a significant jump in the number of prompt credit between Starter to Builder tier). For me, I burn through half of my monthly 250 credit just in 3 days just to finish the front-end part of the solution have not even touch the Backend / API part of my prototype yet — and because you would spend a substantial portion of that on providing correction to some mistake that the Base44 coding agent will, unfortunately, commit.
If you already use other to online vibe-coding platform like Lovable or Firebase, you would appreciate Base44 offer (almost) the same capability
Github repo integration (not available on Starter / Free tier though)
In-chat windows versioning — i.e. you can revert to previous code if you’re not happy with the change (and this action is free for you — which I like)
Latest powerful coding agent: at this time of writing, it give users Claude 4 Sonnet, Gemini 2.5 Pro and (drum-roll) the latest GPT5 (which is a huge step up in agentic coding capability for those who may know)
After a few go, the true picture start to show
I started my project. All well and smooth sailing. Got Authentication and Database sorted right out-of-the-box — which is Base44 main selling points compare to its alternatives. To go a bit further, Base44 use Supabase as their main cloud database vendor — which suit me well as I have used Supabase on my other projects and already familiar with it’s robust interface and APIs system (it’s a must-have Posgresql database for those who don’t want to worry about cloud infrastructure performance, on-going maintenance and their free tier is massively generous so you don’t have to worry about your hosting expense too much until your application start bring in money).
But after a dozen of prompt into and issue slowly emerge:
Github sync is only one-way
I chose Builder because it has offer Github repo integration. However, the bad news is you cannot push your chance the codebase online, and its only allow pulling changes to your local machine.
Of course I can understand the rationale for this restriction is to avoid you accidentally mess up the code on the main repo which would be hard to revert. But personally, I don’t see the value from upgrading to the Builder offers from Starter perspective. What the point if you cannot push your chance to the online repo. And in my case I was the only one using the codebase — what if the project is shared / collaborate between members in a team.
This one-way sync restriction also mean that you have to rely on the Base44 coding agent and cannot utilize your local agent like Claud Code or Gemini CLI to make direct change on the code. The work around for me is to pull the changes to my local repo and use the local coding agent to review and write prompt for the Base44 agent to implement any changes / fixes or making change on their online IDE interface.
Big coding models, mediocre quality
With ability to choose between the Big-3 coding agent (Claude Code, GPT5, and Gemini) you would think that the coding side of things will be easily taken care off. Wrong. After three coding sessions, I was less than being impressed with the code quality the Base44 agent produce — compared to my local Claude Code instance. Even with the latest GPT5, the agent commit silly UI mistake like placing wrong UI icon or missing variables in the code (this was easily seen when the app break and the error details was shown) or logical data connection were poorly executed. So this give me a bit doubt in what actually been used behind the scene as it does not live up to my expectation.
Another example where the AI display a unexpected lowpoint in the agentic coding capability is the initial table schema is poorly designed for my app that need to be catered for multi-tenant usage — a ciritical element allow segregation of context resources between different tenants — i.e. Alice cannot get access to Marcus’s contexts and vice-versa, in run-time and through various states during the user-interaction. I have to explicitly pointed out that so the AI coder to add that in.
Restricted to one tech stack: React SPA run on Vite framework
When I was hitting a technical challenge trying to generate dynamic route for my web app (convolut.app) to replicate the “raw link generation” feature similar to Github — I realize that Vite app does not support dynamic raw link routing as this need to be server-side rendering. I then asked the Base44 AI to rebuild the app into NextJS framework, only to be told that Base44 AI not allowed (or “was instructed not” if that were to be a better words) to build app in other framework other React SPA with Vite framework
