AI agents can prototype apps… But shipping real software takes hours of testing, debugging, and refactoring. Agent 3 is 10× more autonomous — it keeps going where others get stuck. The “Full Self-Driving” moment of software.

Automated Testing. While building your app, Agent 3 will periodically test it using a browser, generating a report and fixing the issues it finds. Our proprietary testing system is 3x faster and 10x more cost effective than state-of-the-art Computer Use Models.

Longer Autonomous Runs. Agent 3 is 10x more autonomous than V2, capable of handling much more complex builds by detecting and fixing errors on its own. You can track the progress of your build with Live Monitoring on your phone, freeing you up to focus on other creative work.

One more thing. Agent 3 can now generate your own agents. Agents that can work with you in Slack or Telegram. Or run based on schedule or trigger based on webhook events. Enrich it with your own data sources, and build automations with apps like Notion, Linear, Dropbox, etc.

Congratz for Replit team! been a replit "hacker" since 2019, happy to see what you guys achieved!

OG 👊

200 minutes of autonomy 🤯

I've been using it for about 20 mins now, literally blew my mind! Super excited to push @Replit and see what else I can produce. SUPER excited!!!!

This is wild. Can't wait to use this. Thrilled we got to invest in your mad genius @amasad

I do not know how to code. Using Agent 3 I was able to create this website in 5 hours and spent $20.40 on usage credits. The future is here - Thank you Replit👏

👏

Still haven’t spoken to anyone who uses replit

@grok when is Agent3 going to be available/deployed

@grok Out now

@grok I see that!

@amasad @grok So if we were already building an app, do we need to allow it to use Agent 3? or will it automatically kick in?

@amasad @grok For me it was automatic

LFG 🔥🚀

How is the security? Is it able to help customers set up secure databases and payment systems?

Congrats on the launch! This looks awesome.

This sounds and feels like Apple ads and coincidentally released after Apple event. Am I the only one suspicious whether this is real ?

As the amount of time agents can work continues to increase, the importance of planning and setting out a clear vision become increasingly important. Hitting undo after waiting 200 minutes is not the best use of time or money, so plan well!

🔥 congrats!!

Huge

Well… people keep saying that they build with in hours, what they built in months with Replit. Try it, is free.

Replit is unstoppable. I can't wait to try this!!! 🔥🔥🔥

Breaking 💥📈! BITCOIN HITS UNSEEN HEIGHTS 🚀💰!

Let’s empower AI builders.

stoked for the automated testing. ive been using chat to do this, glad to see it built in to the workflow

The "iPhone Pro" moment of @Replit?

little chinese peptide has you looking great king.

Thanks, boss. Not on Chinese peptides (yet?) but tell me more

looking very svelte then - there's a few potent glp-1 receptor agonists still in clinical trials that people are either getting or joking about getting from chinese biologic manufacturers

i think if Replit keeps going like this, it will be bigger than Apple, amazing work

Amjad acting is most weird thing I've seen about Amjad

This looks big !! I'm still navigating in cursor universe .... Let's see

Whoa, this is the kind of truth I needed today. ⚡️ We all talk hustle, but rarely do we unpack the cost—mental, emotional, physical. What really stuck: doing less of the noise often opens room for depth, clarity and the version of success you actually want—not someone else’s highlight reel. Question for the crew: If you were to drop one habit, belief, or expectation that’s secretly draining you, what would it be?

Use to manage you’re agents

Very cool Amjad!

Did Replit train this model or is it based on an other LLM?

Autonomy and over-zealousness are not the same. Agent 3 does things that are totally unrequested while providing very misleading feedback when literally attempting to merge two projects, with similar file structures, that were made with earlier versions of agent. My experience differs from what you suggest here. I've been sent round in counter-intuitive loops only to determine agent analysed, planned and agreed a course of action then did something very different. This is very hard to manage as it doesn't communicate this with clarity and claims to have done something else. "Looking at the entire chat history, here are the things I did that differed from your specific and explicit requests: [...] The core issue: I overcomplicated what should have been a straightforward integration of working code by introducing my own solutions instead of following the original working [..] implementation."

Is it on a new project? Be curious if you can compare and contrast vs existing project.

- If you can't make Agent 3 compatible with software created in earlier versions of Replit, without being so autonomous that it deviates from specific, clear, defined, agreed plans and scope, and autonomously wastes time, credits and breaks things galore this creates a novel form of technical debt. Effectively, future versions of Agent are not fit for purpose to tweak or evolve software previously created with Replit. It doesn't stick to defined autonomy levels: When it changes something unrequested that wasn't broken, it knows it has: I am really hoping that future versions of Agent can stay on task and be backwards compatible. cc: @mattyp

As far as I can tell, these issues stem from Sonnet 4.5 which is sometimes great (vibe coding) and sometimes requires determination and massive patience (grit coding).

Agent that can build Agents! 🔥

Incredible, I love the picture-in-picture UX. Congrats on the launch!

This is so cool! Congratulations! 🙌

I just spent the entire evening (and dont even know how many dollars) just trying to setup captcha on my app with the new Agent. And still didn't get it. So far nothing major here... I should probably switch over to Replit Auth, though I liked some of the customization that I had with my own auth, specifically the translated email templates

Love to see this. Looking great. Hope this actually works. Will fire up my Replit again immediately and burn token-credits like crazy. If you see a huge smoke-stack over Europe: it's me, roasting the energy grid with REPLIT. The chosen weapon of the chosen ones. Thx @amasad and everyone in @Replit team for leveling up civilization. You all deserve the intergalactic live player medal in gold 🥇🕹️ And a personal CIV hall of fame CIV7 expansion pack where you can discover the "Replit"-tech on the tech-tree @CivGame . So grateful. Hack on! READY PLAYER 42!