The early-access tool could slash dev time, but its true magic lies in helping indie founders prototype apps quickly and cheaply.
Replit just released Replit Agent v2, the second iteration of the company's AI-powered coding assistant.
Now in early access, the AI agent lets users describe app ideas in natural language — like “create a subscription landing page”— and generates a working prototype.
The announcement, which came from Replit’s VP of AI Michele Catasta, highlighted two big updates from the original version, which launched back in September 2024:
Real-time Design Preview. Users can see how their app will look on the fly — no waiting or guesswork required.
A New App-Creation Experience. Replit redesigned the workflow so founders can brainstorm, pick features, and assemble their code all in one intuitive interface.
Replit’s mobile app, which launched late last year on iOS and Android, supports core features of the AI agent, but hasn't yet added support for Agent v2, according to this tweet from Catasta:
You need a Core ($25/month) or Teams ($40/user/month) subscription to access Replit Agent.
Once you're logged in, you describe the app you want to build, review the AI’s plan, tweak as needed, and deploy.
Usage costs 25 cents per checkpoint (file changes), offset by subscription credits as detailed in Replit’s docs.
I signed up for Replit Agent v2 and played around with it for a while. My opinion? It's cool. But it's not perfect:
Pros: Speeds up prototyping, lowers entry barriers for non-coders, and syncs with Replit’s mobile app for on-the-go work.
Cons: It's limited to new projects started directly in Replit (you can't import an existing app), lots of human oversight is required for complex projects, and subscription costs can add up with heavy use.
I’ve been testing them, and yeah, they’re good. No complaints there. But honestly, if you’re already paying for OpenAI or any other AI, why would you pay for this too? With 3o, you get the exact same results (I wouldn’t be surprised if they’re using OpenAI’s API themselves). So the only real extra value is convenience, which is nice, but worth the price?I’m not so sure.
Yep makes sense. I think we're all dealing with the “tyranny of choice” right now deciding between all these new AI models that do basically the same things.
Replit Agent probably takes the lead for non-technical people who are really intimidated by IDEs and code repositories and all the other trappings of conventional coding environments.
Exactly. It’s true that for non-technical users, Replit is a solid solution, but for now, I think it’s more of a novelty than anything else. From my experience using it, there are moments when the AI hits a wall, it has no idea what to do. But instead of admitting it, it just keeps trying to find a way around the problem. The issue? Every one of those little AI “adventures” costs the user money.
But I don’t want to sound pessimistic. It’s honestly incredible how much progress has been made in such a short time. If anyone should be worried, it’s no-code platforms like Webflow and the rest, because I don’t think we’re far from a future where someone logs into Replit, types, “I want a digital marketing website that looks like this and that,” and boom, two minutes later, they have a Webflow level type of website.
Funny, that's EXACTLY what I am working on...
The workflow I'm shooting for : here are 1000 images of sites. Pick one (or alternatively give me a URL to screenshot) and I'll code that web site for you. Then customize with AI, publish the result wherever.
Biz model: sell tokens ... well, the AI ain't Free, so I was thinking something like 10k tokens for 50 bucks ... Well, something to that effect. You get your web site fast ... and for a pittance (the unused tokens never expire, so you can use them for your next project, whenever that comes to be). No refunds ... too complicated to automate
I agree: Webflow, Wix - that whole category, is on the verge of being disrupted.
I would have sent you guys a link to play with, but I just got this whole idea about 2 hours ago. More realistic (but still crazy - haha) timeframe would be March 1st - 2 days from now. So from idea to MVP: 2 days. If only I could solve my python problem, that's actually feasible. We'll see ...
Delighted to see Replit look out for founders as well. My concern however is that apps built thru Replit by founders would require more supervision than built thru HuTouch, mainly since the generated code from Replit isn't really reliable. Thankfully it's better than no code or template driven coding. Now, HuTouch takes it 10 steps ahead though focusing on generation of > 50% reliable code. HuTouch converts project specs to 1st working version of an app with UI from Figma, functionality from uploaded requirements, API from postman and uses a Developer's coding standards in the process. Would love to give you a sneak peak on HuTouch
This is exciting — tools like this are lowering the barrier to entry for non-tech founders and speeding up MVP development for indie hackers. I’m curious how well Replit v2 handles more complex app logic or custom backend flows.
Anyone here actually built something with it yet? Would love to hear real-world impressions beyond the launch hype.
That’s a fair point — a lot of AI services nowadays do seem to be layering on convenience rather than offering something fundamentally new. If the core model they use is from OpenAI (or a similar provider), you're essentially paying for how they package the experience — smoother integrations, better UI, or extra tools.
The real question is whether those added features save you enough time or effort to justify the cost. If you’re already comfortable using AI directly — like through OpenAI's platform — the value of a middle layer becomes a bit murky.
It’s kind of like paying extra for a fancy calendar app when Google Calendar is free — sometimes the polish is worth it, sometimes it’s not. Would love to know what specifically stood out (or didn’t) for you with 3o!
The topic of recovery services and the efficacy of Wizard Hilton Cyber Tech warrants a closer examination. Recovery services are a critical component of any comprehensive cybersecurity strategy, as they provide the means to restore systems, data, and operations in the event of a breach or other disruptive incident. Wizard Hilton Cyber Tech is a leading provider in this space, offering a suite of advanced recovery solutions designed to ensure business continuity and mitigate the potentially devastating impacts of cyber attacks. At the core of their offerings is a robust, AI-driven platform that continuously monitors systems, detects anomalies, and triggers rapid, automated recovery procedures. This allows organizations to bounce back quickly, often with minimal downtime or data loss. Wizard Hilton’s approach also emphasizes the importance of comprehensive testing and simulation, putting recovery protocols through their paces to validate effectiveness and uncover potential weaknesses. Additionally, their team of seasoned cybersecurity experts provides hands-on guidance, tailoring solutions to the unique needs of each client. By combining cutting-edge technology with deep industry expertise, Wizard Hilton Cyber Tech has established itself as a trusted partner in the realm of recovery services, empowering organizations to safeguard their most valuable digital assets and ensure business resilience in the face of ever-evolving cyber threats. Contact: for assistance,
Email : wizardhiltoncybertech ( @ ) gmail (. ) com
OR
support ( @ ) wizardhiltoncybertech (.) com
WhatsApp number +13024457895
Used it and it was crap!
How is this different from Bolt. new ?
Replit Agent v2 is a great step forward for AI-powered coding, making prototyping faster and lowering barriers for non-coders. The real-time design preview and improved workflow are definite wins. However, its limitations—like no support for existing projects and the need for frequent human oversight—mean it's not a magic bullet for all developers. Plus, the cost could add up for heavy users. Still, for indie hackers and founders looking to quickly test ideas, it’s a promising tool worth exploring!