Describing an app and getting a working one is no longer remarkable. Twelve tools do it, they all demo well, and the demos are the least useful thing about them.
The distinction that matters is not how good the first result looks. It is what you are left holding: a repository you own and can take anywhere, or a working product that lives inside somebody else's platform. Both are legitimate. Choosing without noticing which one you picked is not.
Like our coding agents list, this one carries no prices and no version numbers — both move faster than a page can be corrected. Scores in brackets are this site's own.
Agentic full-stack builders — they write code you own
These run agents that plan, build, test and deploy a real application, and give you the source. If you might one day hand the project to a developer, start here — the export path is the feature.
- Emergent (8.7) — agents that design, code and deploy end to end, producing production-ready code you own and can host anywhere. GitHub integration and private hosting start on the paid tier, and the upper tiers go deep into team and compliance controls
- Lovable (8.1) — the same promise with unusually transparent metering: it publishes what individual actions cost in credits, which makes spend something you can reason about before you act
- Replit (8.4) — build, run and host in one place, with an agent that can take a project from prompt to running service without you leaving the browser. The strongest option if you also want the runtime
- Bolt (7.9) — StackBlitz's prompt-to-app builder, running a full-stack project in the browser and deploying from there. Fastest path from an idea to a URL you can send someone
- Base44 (7.4) — app generation aimed at internal tools and business applications rather than consumer products, where the requirement is usually "works correctly" rather than "looks remarkable"
UI-first generators — they start from the interface
These generate the front end well and leave the rest to you. Useful when you have a backend already, or when the thing you are actually trying to decide is what the product should look like.
- v0 (8.3) — Vercel's generator, producing React components you paste into a real codebase. The best fit when the project already exists and you need a screen, not an app
- Framer (8.1) — a design tool that publishes a real site, with AI generation on top. The pick when the deliverable is a marketing site and design quality is the point
- Uizard (7.2) — turns a sketch or a description into a design you can iterate on. It is a thinking tool for the stage before anyone should be writing code
Visual platforms — they host what they build
You build in their editor and run on their infrastructure. In exchange for giving up the export path you get things the code generators do not hand you: a database, auth, permissions and hosting that already work together.
- Bubble (8.2) — the most capable of the visual builders and the steepest to learn. Genuinely complex applications ship on it, which is not true of most tools in this group
- Glide (7.8) — turns a spreadsheet or database into a usable app quickly. The right answer far more often than people expect, because a large share of internal tools really is a table with rules
- Softr (7.6) — client portals, directories and internal tools on top of data you already keep somewhere else. Fast, narrow, and honest about being narrow
- FlutterFlow (8.0) — visual building for real mobile apps, with code export. The bridge between this group and the first one if the target is an app store
The question that actually decides it
Not "which is best" — ask what happens on day 400.
If the answer involves a developer maintaining this, a security review, or an acquirer looking at it, you need generated code in a repository you control, and you should be on the paid tier of whichever tool you pick, because export paths and version history are usually not free features.
If the answer is that the same person will keep running it and it needs to work rather than scale, a visual platform will get you there faster and keep less of your attention. Paying somebody to host, patch and back up your internal tool is not a compromise — it is the correct trade for most internal tools.
The expensive mistake is building something a business depends on inside a platform you cannot leave, without ever having made that decision consciously.
Five checks before you commit a real product
Run these in an afternoon, before the project is load-bearing rather than after.
- Can you export the code, and on which plan? Test it on day one rather than trusting the marketing page
- Where does the data live, and can you get a backup out without the vendor's help?
- What does a month of your actual usage cost? Run a two-week pilot; published credit figures will not tell you your own burn rate
- What happens when two people need to work on it — is that a supported tier or a shared password?
- Can you point an ordinary developer at it and have them make a change? If not, you have bought a dependency, not an asset