blog-replit-com

BlogNewsAIAnnouncing Replit AI for AllMon, Oct 9, 2023Updated at:Wed, Apr 30, 2025Michele CatastaThe AI Team @ ReplitFor every human endeavor, there inevitably arrives a moment when an innovation profoundly elevates the potential for productivity. For instance, communication was revolutionized by the invention of the printing press, and later, the Internet. After decades of steady progress, software is witnessing its disruption phase as well -- AI is redefining the whole software development lifecycle.AI is one of the key ingredients to go from idea to software, fast -- we witnessed this from over a year of experience in pioneering Ghostwriter Autocomplete and Chat.At Replit, we are on a mission to empower the next billion software creators. Over time, we gained full conviction that such a mission can’t be accomplished without putting our AI features in the hands of every Replit developer.Today, we’re making Replit AI available for everyone. Code completion and AI code assistance are now enabled by default, and available to 23M developers (and counting). Developers on the free plan will have access to our basic AI features, while Pro users will retain exclusive access to the most powerful AI models and advanced features.This was no easy feat, and required not only all our expertise in scaling distributed systems, but also training low-latency LLMs to serve Replit’s compute-heavy AI workloads. Emboldened by the strong performance we obtained with training and serving replit-code-v1-3b, today we are also releasing replit-code-v1.5-3b, a state-of-the-art 3B LLM trained on 1T tokens from a code-heavy pretraining mixture. The dataset includes 30 programming languages and a dev-oriented subset of StackExchange. We will gradually roll out our new code LLM in the next few days, making code completion on the Replit AI code editor even more powerful.Our roadmap for the near future can be summarized in one sentence: AI will redefine every single Replit feature. It won’t be necessary to mention that AI plays a fundamental role every time software is being edited or deployed -- soon we will take it for granted.As part of this, we are saying goodbye to “Ghostwriter” as the name for our AI features. Replit will become a synonym of AI for software creators -- only then we will have accomplished our mission.MoreWed, Mar 11, 2026The Future is Actually Very HumanIn 2016, we started Replit with a simple but radical belief that anyone should be able to build an app without ever learning to code. Our vision was for software to expand human imagination. Instead of requiring people to master software as it existed, we believed software could evolve to meet people where they are, translating their imagination into reality. At the time, this idea felt almost impossible. Software development was for engineers, not students, teachers, designers, or small business owners. The tools were complex, required specialized training, and thus inaccessible to most of the world. The terms vibe coding and AI agents were still nearly a decade away. Fundamentally, we believe the future of technology is deeply human. As software adapts to people, billions will be able to turn their ideas into reality without needing to understand or be restricted by the machinery underneath. AI will unlock human creativity, kicking open a world where far more people can build, experiment, and create than ever before. This vision is fast becoming reality – and we are about to accelerate faster than ever. Replit has raised $400 million, valuing the company at $9 billion, a 3x increase in just six months, reflecting the strength of our business and growth still to come.Tue, Feb 24, 2026Replit Pro Is Here — and Core Now Offers Even Better ValueWe’ve heard from builders that it hasn’t always been obvious which plan is right for them. The truth is, we’re seeing more distinct kinds of builders on Replit than ever before — from people just getting started, to teams shipping serious production projects. That’s why we’re making our plans clearer and more purpose-built. Replit Pro is our most powerful plan yet, built for serious builders who want to push their projects further with the best Agent experience and premium benefits. At the same time, we’re making it easier to get started with Replit Core — with better collaboration features, a lower price, and new controls that help you manage costs. If you’re getting started or building casually, Core is for you. Summary New Pro plan ($100/month). Access to powerful Agent modes, tiered credits with discounts, priority support, up to 15 builders, credit rollover.Wed, Feb 18, 2026Ship Enterprise Data Apps Faster with Replit and DatabricksSummary How Replit and Databricks Apps shorten the path from idea to production data app What Databricks AppKit provides under the hood, and how Replit makes it easier to work with A high-level look at building in Replit and deploying into Databricks Apps Introduction --- Wed, Mar 11, 2026 • FeaturedIntroducing Replit Agent 4: Built for CreativityIntroducing Agent 4 — our fastest, most versatile Agent yet. It's built around a simple idea: you should spend your time creating, not coordinating. Agent 4 takes on the tedious-but-necessary work in the background so you can stay in creative flow and ship production-ready software 10x faster. Because Replit is where software is built, run, and shipped — all in one place — Agent 4 can handle both the complex and the mundane, so you can focus on what's uniquely human: creativity. Summary Agent 4 is built on four pillars designed to keep you in creative flow and ship production-ready apps 10X faster. Design Freely: Generate design variants on an infinite canvas, tweak them visually, and apply the best one directly in your app. Move Faster: Tackle auth, database, back-end functionality and front-end design all at once with parallel agents, with progress across tasks clearly visible. Once done, tasks can be merged seamlessly into the main app. Ship Anything: Create mobile and web apps, landing pages, decks, videos and more in the same project, with shared context and design so you can scale efficiently.AllDesignEventsAIProductEngineeringInfrastructureNewsBuilder SpotlightEduWed, Mar 11, 2026The Future is Actually Very HumanIn 2016, we started Replit with a simple but radical belief that anyone should be able to build an app without ever learning to code. Our vision was for software to expand human imagination. Instead of requiring people to master software as it existed, we believed software could evolve to meet people where they are, translating their imagination into reality. At the time, this idea felt almost impossible. Software development was for engineers, not students, teachers, designers, or small business owners. The tools were complex, required specialized training, and thus inaccessible to most of the world. The terms vibe coding and AI agents were still nearly a decade away. Fundamentally, we believe the future of technology is deeply human. As software adapts to people, billions will be able to turn their ideas into reality without needing to understand or be restricted by the machinery underneath. AI will unlock human creativity, kicking open a world where far more people can build, experiment, and create than ever before. This vision is fast becoming reality – and we are about to accelerate faster than ever. Replit has raised $400 million, valuing the company at $9 billion, a 3x increase in just six months, reflecting the strength of our business and growth still to come.Tue, Feb 24, 2026Replit Pro Is Here — and Core Now Offers Even Better ValueWe’ve heard from builders that it hasn’t always been obvious which plan is right for them. The truth is, we’re seeing more distinct kinds of builders on Replit than ever before — from people just getting started, to teams shipping serious production projects. That’s why we’re making our plans clearer and more purpose-built. Replit Pro is our most powerful plan yet, built for serious builders who want to push their projects further with the best Agent experience and premium benefits. At the same time, we’re making it easier to get started with Replit Core — with better collaboration features, a lower price, and new controls that help you manage costs. If you’re getting started or building casually, Core is for you. Summary New Pro plan ($100/month). Access to powerful Agent modes, tiered credits with discounts, priority support, up to 15 builders, credit rollover.Wed, Feb 18, 2026Ship Enterprise Data Apps Faster with Replit and DatabricksSummary How Replit and Databricks Apps shorten the path from idea to production data app What Databricks AppKit provides under the hood, and how Replit makes it easier to work with A high-level look at building in Replit and deploying into Databricks Apps IntroductionTue, Jan 13, 2026ChatPRD + Replit: From Spec to Running App in MinutesYou draft the PRD. Stakeholders align. Then you wait weeks for development. By the time you see a prototype, requirements have shifted and edge cases emerge. What if your spec became a working app immediately? Product Managers Are Natural Vibe Coders Replit CEO Amjad Masad noted on a recent Reid Hoffman podcast that product managers are some of the best vibe coders. Why? PMs break problems into clear steps and communicate precisely—exactly what AI needs to build effectively. You already know what to build. Replit Agent makes it real. The Integration: PRD to Deployed AppTue, Dec 9, 2025Introducing Race to Revenue to Support the New Generation of BuildersTL;DR: We're giving 20 builders Replit credits, plus an 8-week program, a live CEO Showcase, and strategic perks — with zero equity required. Applications open December 9th. The best businesses being built today don't all follow the same playbook. While venture-backed startups chase unicorn outcomes, there's a parallel universe of builders creating profitable, sustainable businesses. Micro-SaaS products serving niche markets. Bootstrapped companies solving real problems. Solo founders turning side projects into full-time revenue. These builders don't need permission. They don't need to pitch stuffy suits. They just need the right tools, support, and community to accelerate what they're already doing. That's why we're launching Race to Revenue. What is Race to Revenue?Sun, Nov 23, 2025Campus x Replit: Vibe Coding 101 CourseReplit Launches “Vibe Coding 101,” a New Beginner-Friendly Course with Campus.edu Replit and Campus are teaming up to launch Vibe Coding 101—a fresh, beginner-friendly course designed to take anyone from idea to fully built product, no technical background required. Registration is finally open, seats are limited, and the first cohort kicks off this January! Check it out now at campus.edu/vibe-coding Vibe Coding 101 is an immersive 8-week live online program that teaches students how to design, build, and deploy real apps using AI tools. Led by Campus’s world-class faculty, who also teach at institutions like Stanford and Northwestern, alongside industry experts from companies like Meta, PayPal, and Coinbase, the course helps learners ship real projects from week one. By the end, students will walk away with four completed apps, academic credits, and hands-on AI skills they can immediately apply in the real world. The goal: make it possible for anyone, not just engineers, to build useful tools fast. Aligned with Replit’s Core Values Replit is proud to partner with Campus, an institution known for its mission to maximize access to a world-class education. Campus embodies the same values that shape Replit: accessibility, creativity, and collaboration. This partnership is a match made in heaven as both Campus and Replit are committed to making AI upskilling accessible to all. The program also mirrors Replit’s spirit of rapid experimentation. Learners move quickly from concept to prototype, using AI at every step to speed up understanding and creation. Replit’s platform centers on shared learning and community interaction, and the course reflects this ethos through group projects, peer feedback, and open remixing. Students learn together, build together, and amplify each other’s ideas.Tue, Sep 30, 2025Replit acquires OpenIntToday we’re announcing that Replit has acquired OpenInt—our first acquisition. This pairing has a shared history: years ago, I referred the OpenInt team to Y Combinator and they joined the Winter 2023 batch. Since then the team has become a leading integration platform (iPaaS) for code generation. This makes OpenInt a natural product and cultural fit for joining Replit. OpenInt empowers coding agents to write code against any third party service with a focus on correctness and security. OpenInt supports over 160 connectors and now acts as the foundation for Replit Connectors. In contrast to other iPaaS solutions, OpenInt makes it safe for the code generated tools to be created on the fly. This makes it possible for Replit Agent to support the most complex of customer integration needs across the widest breadth of connectors in our industry. Check it out or learn more about it in our dedicated launch post. Thu, Sep 4, 2025 Replit expands Google Cloud Collaboration with new Google Cloud Marketplace ListingWe’re excited to share that Replit is now available for purchase on Google Cloud Marketplace—marking another key milestone in our collaboration with Google. Replit’s Enterprise customers already have their apps running on Google Cloud’s world-class infrastructure, giving every app the security and scalability of Google Cloud. For example: Hosting & Deployment – Replit app deployments run on Google Cloud Storage – Replit App Storage is powered by Google Cloud Storage Security – Apps are protected against DDoS attacks with Google Cloud Armor Encryption – Secrets are safeguarded using Google Cloud’s secure storageThu, Feb 27, 2025Introducing WorkflowsWith Workflows, you can easily save your development processes and organize them as a sequence of Tasks. Instead of having to repetitively type commands into your shell and remember them, Workflows provide an intuitive interface to configure your frequently used commands and execute them with the click of a button. Simplifying development with Workflows Workflows give you the ability to better organize your project execution pipeline, while offering an accessible interface to reuse them. Using the Workflows pane, you can easily configure specific buttons to run specific commands, and even link multiple Workflows together in a chain. Instead of copy pasting "cd ./client && npm run dev && cd ../server && node server.js", you can define the procedure as a Workflow with multiple Tasks. Let's see how it works with an example. Wed, Feb 26, 2025Replit's All-new Database EditorThis week, we launched a new Database Editor feature on Replit. This feature is another addition to Replit’s full suite of tools that make it easy for anyone to build an app — no coding knowledge required. When you build an app on Replit, you get a fully-functioning project that’s ready to send out to the world. This includes real databases powered by Postgres: the most popular database in the world, that powers many of the applications you use every day. Did you know? Postgres is used by half of all professional software developers, according to StackOverflow’s developer survey. Replit handles all the setup and management of your databases, so it’s never been easier to get started. Now, we’re making it even easier to use databases on Replit. Inside your app’s workspace, head to the “database” pane, and click to the “My Data” tab. Here, you’ll see a spreadsheet-like view of your app’s data.Tue, Jul 16, 2024Introducing Replit TeamsIn today's fast-paced work environment, using real-time collaboration tools like Google Docs and Figma comes naturally to most of us. As a result, we've come to expect a lot more from our tools. Yet, when it comes to building software, we often find ourselves stuck with tools built for a different era — tools that cause our teams to feel siloed, inefficient, and less collaborative overall. In decades past, the software development lifecycle looked like a rigid industrial process: product managers would hand off documents to designers, and designers would hand off mock-ups to engineers. Today, however, the process is much more fluid and cross-functional. Engineers contribute to design, product managers prototype, and designers code. Still, engineering tools have remained inaccessible to everyone except engineers, preventing meaningful contributions from other roles, and slowing everyone down. This led us to wonder: What would it look like to design software collaboration tools that truly meet the changing needs of 21st-century teams? Introducing Replit Teams Go from idea to software faster than ever before. Today, we’re announcing that Replit Teams is generally available and here to set a new standard for building software. Wed, Dec 27, 2023Replit in Review: A Recap of What We Shipped in 2023In 2023, we brought the vision of idea to software, fast, to reality. We shipped the first conversational AI programmer to code your ideas faster, made AI available for everyone, and introduced Deployments to ship from where you write your code. We made coding on Replit a better and more productive experience by redesigning the Console, Git integration, Packages tool, and more. We released the Replit Desktop App so you could choose the form factor that works best for you. We’re not finished making Replit the best place to code, but here are some highlights from the past year! Product (01/29) Git in the Shell - The first step in improving our Git integration. Git command-line tools were improved and integrated with your GitHub account, allowing access to your repositories. (02/14) Announcing Ghostwriter Chat - The first conversational AI programmer. Ghostwriter Chat is a beta release that acts as a proactive debugger and understands your project’s context, making it a powerful AI pair programmer.Tue, Dec 26, 2023Introducing Spellburst, A Replit x Stanford HAI CollaborationAt Replit, we’re lucky to work with brilliant people who are building the programming tools of the future. This past year, we had the privilege of sharing some of this design work with the broader academic research community. Tyler, design engineer at Replit, teamed up with Stanford University (Jenny Han – also former Replit intern), Miroslav Suzara, Christopher Pondoc, and Dr. Hari Subramonyam) to publish Replit’s first-ever peer-reviewed publication, Spellburst. In November, we presented our work at the ACM Symposium on User Interface Software and Technology (UIST), one of the premier venues for human-computer interaction research in the world. This post will walk through what we built and researched, what we learned from the process, and some predictions for user interface design and AI in the future. Introducing Spellburst We went into this project knowing that we wanted to design for learning and creativity for coders. In our formative study, we interviewed 10 expert generative artists to understand their existing workflows. What we discovered about the exploratory process for creative work is that creative coders are constantly: Going from idea to codeMon, Nov 20, 2023Transitioning to Replit CoreLast week, we launched Replit Core - a singular membership plan that represents the end-to-end software builder experience. We heard a lot of excitement and questions, so we wanted to clarify Replit’s new membership model and what that means for Replit moving forward. We value our existing community that grew Replit to be the world's largest community of software builders, where anyone can create and learn how to go from idea to software. Along with last week's announcement, we wanted to honor your commitment to us by letting you keep your current monthly price. You will be upgraded to Replit Core for the remainder of your current subscription term. At this time, if you are on auto-renew, the next renewal will be at your current Hacker rate. In addition, members with a monthly Pro plan who would like to upgrade to Replit Core with our promotional pricing can now do so by visiting replit.com/pricing and selecting ‘Switch to yearly payment ($120)’. This will automatically cancel their current plan and switch them to the annual plan at the promotional rate. We’re on a mission to empower the next billion software creators and we’re excited about bringing the best of Replit under a unified membership plan. We firmly believe that $20 per month / $220 per year represents the best value in software development, and we aim to continue building on that value with your feedback in mind. For more information, check out the breakdown. Over the next few months, we will explore enhancements to the Replit workspace, AI, and Deployments improvements. We will pursue partnerships with other platforms to bring Replit members even more value. One of these perks is available today where Replit Core members enjoy a complimentary 4-month subscription to Perplexity Pro. Thanks for being a part of our builder journey as we continue to support yours.Tue, Nov 14, 2023Announcing Replit Core - The Essential Membership for BuildersAt Replit, we are on a mission to empower the next billion software creators. As we reflected on our mission and heard feedback from our community, we grew confident in our pursuit to create an end-to-end experience that represents the best developer tool membership in the market to go from idea to software, fast. Today, we’re excited to announce Replit Core, a singular plan that we aim to continue investing in for years to come. With Replit Core, you get: Access to advanced Replit AI features Save time and code faster with Replit's industry-leading AI coding assistant, currently powered by GPT-4. Debug code, get intelligent autocomplete suggestions, and convert natural language to code with one click. 12345 --- Wed, Mar 11, 2026 • FeaturedIntroducing Replit Agent 4: Built for CreativityIntroducing Agent 4 — our fastest, most versatile Agent yet. It's built around a simple idea: you should spend your time creating, not coordinating. Agent 4 takes on the tedious-but-necessary work in the background so you can stay in creative flow and ship production-ready software 10x faster. Because Replit is where software is built, run, and shipped — all in one place — Agent 4 can handle both the complex and the mundane, so you can focus on what's uniquely human: creativity. Summary Agent 4 is built on four pillars designed to keep you in creative flow and ship production-ready apps 10X faster. Design Freely: Generate design variants on an infinite canvas, tweak them visually, and apply the best one directly in your app. Move Faster: Tackle auth, database, back-end functionality and front-end design all at once with parallel agents, with progress across tasks clearly visible. Once done, tasks can be merged seamlessly into the main app. Ship Anything: Create mobile and web apps, landing pages, decks, videos and more in the same project, with shared context and design so you can scale efficiently.AllDesignEventsAIProductEngineeringInfrastructureNewsBuilder SpotlightEduFri, Feb 13, 2026Vibe Coding Data Apps with Replit + Snowflake - Part 2 [webinar]Summary Snowflake’s Vino Duraisamy and Replit’s Manny Bernabe return to turn simple dashboards into working data tools. Live forecasting and drilldowns explain why metrics move, not just what changed on your Snowflake dashboard. Custom Slack alerts fire when revenue crosses key thresholds so teams catch spikes, drops, and odd days quickly. An embedded AI data assistant answers plain-English questions and shows the exact SQL it ran for every query. Replit handles security scanning, auth, and hosting so you can ship Snowflake-powered data apps in minutes.Fri, Jan 30, 2026Vibe Coding Data Apps with Replit + Snowflake [webinar]Summary Snowflake developer advocate Vino Duraisamy joins Replit Community Manager Manny Bernabe for another informative live webinar. Replit's Snowflake connector enables builders to vibe code custom data applications without writing code manually. The connector provides secure, read-only access to Snowflake databases, eliminating complex authentication and setup processes. Builders can iterate on dashboards and reports using natural language prompts in plan and build modes. Publishing production-ready apps takes minutes with built-in security scanning, authentication, and deployment features all included.Fri, Jan 23, 2026Building a Private Bookmark Manager and Browser Extension in Replit [webinar]Summary Replit hosted a livestream build session showing how to create a private bookmark manager with a Chrome extension from scratch using Replit Agent in under two hours. The team demonstrated how to write effective prompts, troubleshoot issues, and add AI-powered features like content summarization using OpenAI integration with minimal manual coding. Building the initial functional app and extension cost around five dollars using Replit Agent, with additional costs for feature iterations and enhancements throughout the session. The livestream showcased real-world debugging and problem-solving, including handling database migrations, security vulnerabilities, and browser extension authentication flows as they occurred. The start of 2026 is proving to be an exciting time for builders on Replit. In a recent livestream, community managers Manny Bernable and Raymmnar walked over 1,000 viewers through building a complete bookmark management application with a Chrome extension—live and unscripted—using Replit Agent.Wed, Jan 14, 2026How RevOps Teams Are Building Their Own Tools with Vibe CodingSummary RevOps professionals are using Replit to build custom micro tools that solve niche workflow problems without technical backgrounds. Teams save 90% of their time by automating tedious processes like mock data generation, CSV enrichment, and HubSpot syncing. Sales teams use AI-powered training simulators to practice cold calls and discovery calls, reducing ramp time from months to weeks. Vibe coding eliminates dependency on engineering resources, allowing go-to-market teams to iterate on solutions in hours instead of months. Building custom tools is becoming a core competency across sales, marketing, and operations—not just for developers anymore.Wed, Sep 3, 2025Introducing Plan Mode: A safer way to vibe codeWe are excited to announce Plan Mode, a new way to collaborate with Replit Agent. This mode provides a dedicated space for ideation, task outlining, and strategic discussions—all without initiating any changes to your codebase. The default mode is now referred to as Build mode. Key benefits Plan Mode allows you to: Brainstorm ideas and explore different approaches to your projectTue, Apr 22, 2025Secure Vibe Coding: The Tools and Fundamentals to Vibe Code SecurelyNothing is more exciting than realizing your vision. Nothing is worse than realizing you had a major security flaw. The good news? While security is crucial, it doesn't have to be an obstacle, especially when you're building on Replit. We integrate many security features directly into the platform, and our AI tools can handle much of the heavy lifting, letting you focus on building. This post is your practical guide to navigating security essentials. We'll walk through the key fundamentals to secure vibe coding, provide a security checklis, and give 16 practical tips for how you can vibe code more secure apps (and how to do so simply). Replit's five fundamentals for secure vibe coding Up to 40% of AI suggestions may contain vulnerabilities. At Replit, we've taken a unique approach: making security automatic and integrated for the most secure vibe coding possible.Wed, Mar 26, 2025Secure Vibe Coding Made SimpleAs vibe coding revolutionizes app development, security can't be an afterthought—it needs to be a core part of the creation process. Recent Stanford research shows that developers using AI assistants wrote less secure code in 80% of tasks compared to those coding manually, while being 3.5 times more likely to overestimate their code's security. And that's professional developers, we've seen an even more concerning trend among vibe coded apps. At Replit, we don't just take this seriously, we build security into every app you create with Agent, by default—something Replit CEO Amjad Masad has called "the Pit of Success". That's why our apps are secure, by default. We bake security in to the process of building with Agent—so you can focus on creating.Wed, Mar 26, 2025What is Vibe Coding?From idea to app, fast—that's always been our mission at Replit. But what if you could turn your ideas into working apps without knowing how to code at all? Enter vibe coding: a term that's democratizing app creation for everyone, regardless of technical background. What is vibe coding? Vibe coding refers to the practice of instructing AI agents to write code based on natural language prompts. It's not about being lazy—it's about focusing your time and energy on the creative aspects of app development rather than getting stuck in technical details.Thu, Mar 13, 2025Agent Just Got Eyes — Click. Describe. Done.Making changes to your app should be as fast as possible. Instead of manually describing what to edit, what if you could just click? Now you can. Meet Element Selector, a simple yet powerful way to select any UI element in your app and modify it instantly with Agent or Assistant. No need to type out descriptions or search through files — just Click. Describe. Done. How It Works Toggle the Element Selector — You can find it alongside the chat input tools in Agent and Assistant. Click to select — Want to edit a button? Change a heading? Just click it. Describe changes — Assistant updates text, colors, or styles in seconds. Use Agent to add features like turning a static list into a searchable one.Sat, Mar 8, 2025Everything you need to know about MCPYou can find the 3-min quickstart from the video here What is MCP? MCP (Model Context Protocol) is a standard way to connect AI models to data sources and tools. It allows AI to access information and capabilities beyond what they were originally trained on. MCP works like a universal connector for AI systems. Just as standard ports let different devices work together, MCP lets different AI models connect to the same tools and data sources. This means developers can build a tool once and have it work with any AI model that supports MCP. It's like the USB-C port of AI applications.Tue, Feb 25, 2025Introducing Replit Agent v2 in Early AccessSmarter with realtime app design preview In partnership with Anthropic’s Claude 3.7 Sonnet launch, we're excited to announce the release of Replit Agent v2 in our early access program. The early access program is available exclusively to users on Replit paid plans who have opted in to Explorer Mode (for instructions on opting in to Explorer Mode, see below). While we're encouraged by initial results, you may encounter speed bumps and areas needing improvement as we continue development. What's new: A more autonomous AgentTue, Feb 4, 2025Replit launches the first software creation Agent on iOS and AndroidYou've always wondered why that app didn't exist. Now you can — Make an App for That. We’re thrilled to announce that everyone can now try Replit Agent for free! Check out what we’re launching: Replit launches the first software creation agent on iOS and Android — use our AI app builder to build and deploy apps from your iPhone or Android. Chat with Agent as if texting a friend, and watch as it writes, deploys, and hosts your app in real time — no laptop required. Free Tier for Everyone: Replit Agent is now accessible to everyone with a free tier that includes trying Agent, making it easier than ever to start building apps!Thu, Jan 16, 2025AI Coding is Better with Python"Can everyone really code with AI?" I hear this question all the time. Usually after someone sees a viral post about an AI-built app and thinks "I could do that too!" But here's what those posts don't tell you: most AI-generated apps are just pretty shells—nice looking websites that lack functionality. Why? Most AI coding assistants build with JavaScript and are optimized for presenting information to the user (frontend development). These apps lack the backends and services that power functional applications. JavaScript is great. It creates beautiful frontends—Replit’s AI tools use it well. But Replit is unique: we don’t only use JavaScript. In fact, there are many times it’s not the right tool for the job. Perhaps you’re:Tue, Dec 10, 2024Announcing the New Replit AssistantToday on Replit, anyone can take their ideas and turn them into software — no coding required. For nearly a decade, Replit’s mission has been to bring coding to the next billion people. We’ve helped millions of people learn to code and made programming less intimidating. But now, thanks to our Agent and Assistant, our AI coding products, we’re taking a giant leap forward. You don’t need to learn coding to be a creator — you just need an idea. From local realtors in Cleveland to Japanese influencers to product managers at some of the world’s largest enterprises, people are building incredible things on Replit. We no longer need to limit ourselves to being users of software, we can all be creators. Today, we’re taking Replit Agent out of early access and officially launching it with groundbreaking updates that will radically expand who can create software. Let us show you what’s possible.Tue, Oct 8, 2024From localhost to liveIf you have a body, you are an athlete. Bill Bowerman, Co-founder, Nike The title "programmer" was always reserved for those who code. So it make sense that shipping software is synonymous with, well... writing code. But is the hardest part about shipping really learning to code? I wish. The hardest part(s) are everything that lives between localhost and live. Today, we're seeing a new class of builders—the AI native developer. These users have some technical prowess and build through clever use of the latest tools. But the problem was never just writing code. Despite lower friction to code, roadblocks still exist. Where, you ask?123 --- BlogProductAIGhostwriter AI & Complete Code BetaWed, Sep 7, 2022Updated at:Tue, Oct 17, 2023Amjad MasadSamip DahalAlexandre Cai Giuseppe BurtiniUpdate: Ghostwriter is out now!Activate GhostwriterIn 2018 when we announced Multiplayer Mode, we said it's the most significant evolution of Replit to date. For the first time, you could share a URL with a friend, student, or coworker and get a shared text editor and runtime — no setup required. Replit Multiplayer is changing how an entire generation of programmers learn how to code and make software.Today, we're announcing Ghostwriter, which infuses state-of-the-art intelligence into nearly all IDE features. Ghostwriter sports an ML-powered pair programmer that completes your code in realtime, tools to generate, transform, and explain code, and an in-editor search utility that lets you find and import open-source code without leaving your editor (think Stackoverflow in your editor).Ghostwriter is like Multiplayer in that you collaborate in real-time with someone else. However, in this case, you're not coding with a person; instead, it's an agent representing the entire programming knowledge of the human race. We believe Ghostwriter will leapfrog traditional IDE features. Ghostwriter is the next major evolution of our platform. We think this will radically change how people write code on Replit — and in the process, will change software writ large. Forever.Complete CodeGhostwriter's flagship feature is Complete Code: an AI-powered pair programmer. We believe that Ghostwriter Complete Code is faster, more powerful, and more accessible than any other comparable offering. The best thing about Ghostwriter? It makes writing code on mobile devices not only tolerable, but actually enjoyable: Swipe right to accept!Ghostwriter's Complete Code is in closed beta right now. Please sign up here if you'd like to help us test it. Here is what alpha users are saying about it:"The first thing me and all my friends noticed was how much faster it is than GitHub Copilot. It is at least 2x faster, maybe 3x. It's a little detail but it makes a big difference.""It makes web development so much easier. I feel like I'm only writing 50% of the code.""After using the feature for only week, I can't imagine life without it""It's crazy how much faster I can learn new things without leaving the editors"Building Ghostwriter Complete CodeWhat do you do when you're not a multi-trillion multi-national corporation (yet) with tons of ML research scientists, infinite budget for training, billions in industry partnerships, and store most of the world's code but still want to bring state-of-the-art AI to production? You start from open-source!Open-source Large Language Models (LLMs) like Salesforce's CodeGen models are a fantastic place to start. However, with billions of parameters, they are insanely high latency off-the-shelf. This is especially a problem for code completion, because the models need to be fast enough not to disrupt the user's flow. To be able to serve millions of users, we've been playing with a few optimization techniques to achieve super low latency at a reasonable cost. And we've made tremendous strides in a short amount of time — our median response time today is less than 400ms, which makes our product the fastest in the world.Here we'll go over some of the optimization tactics we've used or are actively exploring:FasterTransformer + Triton server:First, we convert CodeGen checkpoint to FasterTransformer to make use of their highly-optimized decoder blocks (with the possibility of extending to distributed, multi-GPU fashion). Their computational and memory optimizations make inference much faster than popular frameworks like PyTorch or Tensorflow. On top of this, we use Triton, Nvidia's inference server, which is super fast and scalable.Distillation:For further speedup, we perform knowledge distillation of CodeGen model with 2B parameters down to a lightweight and fast student model with roughly 1B parameters. The student model is trained to reproduce the larger model closely while having far fewer parameters and being less computationally expensive. Hence, it is more practical when operating at scale.Quantization:We are also exploring post-training quantization of weights and activations to int8 precision with quantization-optimized kernels to bring down latency. Performing computation in such low precision with optimized kernels often improves latency without causing a significant loss in the accuracy of the model.Future improvementsWork is already underway to improve Ghostwriter further:Further training on open-source datasets like CodeParrot.Deep Reinforcement Learning to further train LLMs with additional signals like user feedback, accuracy on unit tests, compiler/runtime errors etc.Humans don't code unidirectionally. They go back and forth to add, delete, and edit code. Autoregressive language models generate code in one forward direction. Recent works have made LMs more flexible by infilling training. They divide code blocks into , mask middle and train the model to predict middle given prefix and suffix. Because Replit's IDE stores operational transformation edits, we capture the natural cursor movements of human programmers and their code edits, which contains much richer information than synthetically made infilling datasets. We plan to train LMs to predict the OT distribution. We think it will make LMs much more of a pair programmer.Editor implementationGetting the model right is only half the battle. Surprisingly, the client-side implementation is equally as challenging as training and running the models.The user experience for any AI application is paramount to making it feel helpful (instead of annoying). The nitpicky detail necessary to get this right is immense; here is a sampling of some of the issues we have been working on.Whitespace and brace matchingWhen the model generates a recommendation, it might contain code that already exists in the surrounding context. The simplest form of this is with whitespace and braces. A detail that seems tiny on the surface but matters a lot for a flawless user experience.So, to fully understand this, imagine the following (incomplete) code:const isOdd = useCallback(function isOdd(n) { | }, []); You’ve stopped, with your cursor tabbed in to where the vertical bar is, and are about to receive a suggestion from the model. The model sends back its recommended completion: return n % 2 === 1; }, []); This includes a leading tab and a trailing brace, both of which already exist, so if we just blindly dumped the recommendation into your code, you’d be left with:const isOdd = useCallback(function isOdd(n) { return n % 2 === 1; }, []); }, []); A decent recommendation turned frustrating by an incomplete user experience.Instead, we match and filter on certain matching whitespace, abstract syntax tree (AST) characteristics to produce the desired recommendation:const isOdd = useCallback(function isOdd(n) { return n % 2 === 1; }, []); Showing the correct code recommendation is only half of the equation. Very often, users keep typing even after the code recommendation is shown. If you type something that matches the suggestion, we need to adjust the suggestion shown on screen to hide the part the user has typed—and resurface it when you hit backspace.What's more, when you type, the editor would autocomplete a beginning bracket whose corresponding ending bracket; we want to make sure the latter appears in the right spot. As an example, suppose you have typed// Merge [number, number] with [string, string] const mergeTuples = (first: |) and the suggestion isconst mergeTuples = (first: Array, second: Array) => { We want to make sure code recommendations do not disrupt existing code, and correctly splits itself into the part before ) and after it.uxHeuristic filteringA common challenge with LLMs is that at times, they can generate useless suggestions, annoying repetition, or things that are completely wrong.To produce something that actually feels like “intelligence” requires a bit more sophistication. We apply a collection of heuristic filters to decide to discard, truncate or otherwise transform some suggestions; soon, we’ll also apply a reinforcement learning layer to understand the kinds of suggestion that are helpful to users, filtering out suggestions that are unlikely to be accepted to prioritize suggestions that are genuinely helpful.StreamingWe can't overemphasize how much speed matters here. Anyone who has used a sluggish IDE knows how frustrating it is, and outdated suggestions that can get in your way will easily make this feature a net negative on the user experience.In addition to all the model optimizations we detailed above, we also implemented streaming. In other words, we don't have to wait for the entire recommendation to be available, so we literally just start presenting the generated code as soon as possible, chunking it into your UI line-by-line as it becomes available.This little detail makes an enormous difference to how fast the AI feels and how easily integrated into your actual programming experience it is. Maybe it also feels a bit more “intelligent” to know it takes time for the computer to figure out exactly how it wants to help.A Society of ModelsOur conceptual model for Ghostwriter that it's a pair programming agent. It's tempting to map this to a single model that runs all the features. However, as we made progress, we realized that it's better to think of Ghostwriter as a society of models of different shapes and sizes helping you succeed.Semantic SearchA large amount of code exists in open source, but it’s hard to search with natural language because natural language and code are two very different modalities. So instead of deploying traditional approaches like keyword matching, we use embeddings from transformer-based models to power code search. Specifically, we use a finetuned version of CodeBERT model to get learned representations for code and query. The CodeBERT model is finetuned to map both code and query to vectors in joint vector space that are close to each other. We then conduct nearest neighbor matching between the code and query vectors/representations. Such learned representations of code can encode information about what the code does, in addition to other characteristics like keywords the code has, etc. Hence, during inference, the user can search for code in plain natural language by specifying what the code should do.Importantly, users can search for code from inside the editor. This allows us to improve code search even further by making search contextual. Meaning, we give the ML model access to the code the user has already written, whenever searching for upcoming code. This allows us to exploit the clues present in user’s code (like libraries being used) that makes search tailored to that user’s context. We achieve such contextual code search by training CodeBERT model to minimize the distance between single embedding of code context + query, and code. More details can be found in this paper.searchExplain CodeLarge models are especially good at reasoning tasks, and explaining what a piece of code does benefit from every last bit of model performance we can get. For our Explain Code feature, we use the largest state-of-the-art code models, in this case powered by OpenAI.Generate CodeWhile Complete Code is super useful for interactive experiences, sometimes users are willing to take a pause in order to generate entire programs or files. That kind of generations benefits from the performance of insanely large models (100B+ parameters).generateTransform CodeFinally, we leverage exceptionally large models to provide prompt-driven refactor/rewrite experiences.transformThe futureWith the advent of LLMs and generative models in general, we believe that software is entering a new epoch. In the near future, anyone with time and good ideas will be able to build amazing things. AI will guide you as you learn new concept, push just-in-time useful information to you, and even comment on and critique your code. This brings us much closer to our vision of bringing the next billion software creators online, and in the process reducing the distance between ideas and wealth.Many of the Ghostwriter features are already available for Hacker subscribers, and more are coming. Ghostwriter Complete Code will be in closed beta for the next few months as we continue to make improvements to it. If you're interested in trying out, please sign up here.Over the next few months we'll be packaging up Ghostwriter into a Cycles-based power-up that anyone can buy. We're hoping to make this feature more affordable than other offerings on the market. Eventually, however, the plan is — just like Multiplayer mode — to make Replit AI-powered by default and freely available to everyone.Have some questions? See our FAQ document.MoreMon, Mar 23, 2026Live from Replit HQ Part 2Summary This post recaps part 2 of our "Live from HQ" livestream series covering the Agent 4 launch. Peter engineered Agent 4's parallel task system, enabling AI to resolve merge conflicts automatically 90% of the time. Haya Odeh, co-founder and design lead, built the Infinite Canvas to close the gap between design and engineering. Jacob's collaboration features give builders real-time visibility into who's working on what inside a shared project. Adi made branching feel instant using micro VMs, creating isolated task environments that spin up in seconds.Fri, Mar 20, 2026Live from Replit HQ: Agent 4 Launch Pt. 1Summary Manny and Raymmar opened the doors at HQ for a two-hour behind-the-scenes look at our team, the energy after launch, and what Agent 4 actually means for builders. Manny demoed a taste-development app built in Agent 4 with a landing page, web app, and mobile-native version all inside one unified project on the Infinite Canvas. Raymmar showed Replitopolis, a live 3D city pulling from our BigQuery data where every building is a real user and every orange glow is a commit. Eight other team members joined the stream, from designers and engineers to our marketing team and Amjad, our CEO and co-founder. Last week was Agent 4 launch week, and the team could finally breathe. Our very own community managers Manny Bernabe and Raymmar Tirado decided the best way to mark it was to bring the community inside HQ: pull teammates on camera, share what everyone has been building, and have fun along the way.Wed, Mar 11, 2026Introducing Replit Agent 4: Built for CreativityIntroducing Agent 4 — our fastest, most versatile Agent yet. It's built around a simple idea: you should spend your time creating, not coordinating. Agent 4 takes on the tedious-but-necessary work in the background so you can stay in creative flow and ship production-ready software 10x faster. Because Replit is where software is built, run, and shipped — all in one place — Agent 4 can handle both the complex and the mundane, so you can focus on what's uniquely human: creativity. Summary Agent 4 is built on four pillars designed to keep you in creative flow and ship production-ready apps 10X faster. Design Freely: Generate design variants on an infinite canvas, tweak them visually, and apply the best one directly in your app. Move Faster: Tackle auth, database, back-end functionality and front-end design all at once with parallel agents, with progress across tasks clearly visible. Once done, tasks can be merged seamlessly into the main app. Ship Anything: Create mobile and web apps, landing pages, decks, videos and more in the same project, with shared context and design so you can scale efficiently.