Bolt.new
Last reviewed
May 17, 2026
Sources
33 citations
Review status
Source-backed
Revision
v8 ยท 7,214 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
May 17, 2026
Sources
33 citations
Review status
Source-backed
Revision
v8 ยท 7,214 words
Add missing citations, update stale details, or suggest a clearer explanation.
Bolt.new is an AI-powered full-stack web development platform built by StackBlitz. Launched on October 4, 2024, Bolt.new allows users to describe an application in natural language and receive a complete, running web app inside the browser, with no local development environment required. The platform combines frontier large language models (initially Claude 3.5 Sonnet by Anthropic) with StackBlitz's WebContainers technology to generate, execute, and deploy full-stack applications entirely within a browser tab [1].
Bolt.new's growth following its launch was extraordinary by any measure. The product went from zero to $4 million in annualized recurring revenue within 30 days, reached $20 million ARR in roughly 60 days, and hit $40 million ARR within five months [2]. By December 2025, StackBlitz had grown the platform to over 7 million users globally, and by mid-2026 the company reported that 75% of the Fortune 500 had adopted Bolt.new in some capacity [21][22]. This trajectory made it one of the fastest-growing software products in history, second only to ChatGPT, and placed it at the center of the emerging "AI app builder" category alongside competitors like Lovable, Replit, and v0 by Vercel.
StackBlitz was founded in 2017 by Eric Simons (CEO) and Albert Pai (CTO), both Chicago natives who met as teenagers and began coding together. Before StackBlitz, Simons had a colorful entrepreneurial history. In 2011, he co-founded ClassConnect, an education startup accepted into the Imagine K12 accelerator. When funding dried up, Simons famously lived secretly inside AOL's Palo Alto headquarters for nearly two months, sleeping in the office, showering in the company gym, and surviving on free cafeteria snacks while continuing to code. Security guards assumed he was simply working late until he was eventually discovered and escorted out [3]. After ClassConnect, Simons and Pai built Thinkster in 2013, an online coding education platform that was acquired in January 2019.
StackBlitz launched as a browser-based integrated development environment (IDE) built on top of Visual Studio Code. The company's core technical differentiator was its work on WebContainers, a browser-based runtime that enables Node.js applications to execute entirely inside a browser tab. The team spent over four years developing this infrastructure before it became the foundation of Bolt.new.
WebContainers are a micro-operating system built on WebAssembly that can run Node.js, npm, and related tooling natively inside a browser tab. The technology was first introduced publicly in 2021 and represented a fundamental breakthrough in browser-based development. Rather than relying on remote servers or cloud virtual machines, WebContainers execute all code directly on the user's device within the browser's security sandbox [4].
The implementation relies on several key web technologies working together:
| Component | Role | Description |
|---|---|---|
| WebAssembly | Core runtime | Powers a micro-operating system capable of running Node.js entirely in the browser |
| Service Workers | Network layer | Provides a virtualized TCP network stack mapped to the browser's ServiceWorker API, enabling local development servers |
| SharedArrayBuffer | Memory sharing | Allows simultaneous memory access from multiple workers, critical for process coordination |
| Cross-origin isolation | Security | Ensures the sandboxed environment operates safely within the browser's security model |
WebContainers deliver several performance advantages over traditional cloud-based development environments:
Browser support for WebContainers includes full support in Chrome and Chromium-based browsers, with beta support in Firefox (since version 104) and Safari (since version 16.4). Partial support is available on Android browsers and iOS/iPadOS Safari [5].
The WebContainers API was made publicly available for third-party developers, enabling other companies and projects to embed browser-based Node.js environments in their own applications. Common use cases beyond Bolt.new include interactive programming tutorials, next-generation documentation with live code examples, browser-based IDEs, and employee onboarding platforms.
Despite the technical achievement of WebContainers, StackBlitz struggled to find significant commercial traction as a standalone IDE product. By late 2023, the company's annual revenue was approximately $80,000 with flat growth. At a December 2023 board meeting, investors delivered an ultimatum: demonstrate real traction in 2024 or wind down the company [2].
The team began exploring how AI could transform the WebContainers platform. In February 2024, they built an initial prototype combining AI code generation with browser-based execution, but the results were underwhelming. The AI models available at the time could generate code snippets but consistently failed when tasked with producing complete, functioning applications. The prototype was shelved.
The pivotal moment came in June 2024 when StackBlitz gained early access to Anthropic's Claude 3.5 Sonnet. Eric Simons recognized that Claude 3.5 Sonnet's code generation quality was qualitatively different from previous models and that combining it with WebContainers could create something entirely new: a system where an AI model generates an entire application and a browser-based runtime executes it immediately, allowing the user to see results in seconds [2].
Simons built the working prototype of what would become Bolt.new in a matter of weeks. The concept was simple but powerful: the user types a prompt describing what they want to build, the AI generates the complete codebase, WebContainers install dependencies and run the application in the browser, and the user sees a live preview alongside the generated code.
Bolt.new launched publicly on October 4, 2024, announced through a single tweet. The product gained immediate traction, driven in large part by viral demonstrations on social media where users showed complete applications being generated from short text prompts. On its first day, Bolt.new generated $60,000 in ARR; by day two, that figure had risen to $80,000 [6]. Within the first few weeks, active paying customers surged from roughly 600 to over 14,000 [2].
The growth was so rapid that it temporarily maxed out Anthropic's GPU capacity. Anthropic CEO Dario Amodei reportedly described Bolt.new as the fastest-growing customer they had ever seen [2]. By December 2024, Bolt.new had reached approximately $20 million in ARR with just 35 employees.
In January 2025, StackBlitz raised $105.5 million in a Series B funding round led by Emergence Capital and GV (formerly Google Ventures), with participation from Madrona Venture Group, Conviction, and Mantis (the venture fund of The Chainsmokers). Joe Floyd of Emergence Capital joined the board. The round valued the company at approximately $700 million [7]. This came on top of the company's $7.9 million seed round raised in April 2022, led by Greylock Partners with support from GV, Tom Preston-Werner (GitHub co-founder), and others [8].
By March 2025, Bolt.new had reached $40 million ARR, had powered over one million websites in partnership with Netlify, and had accumulated over 3 million registered users with 1 million monthly active users [1][6]. By mid-2025, the platform claimed over 5 million total users, adding more than a million new users per month, and by December 2025 the user base had grown to more than 7 million globally [21]. Analysts projected that the company would cross the $80 million to $100 million ARR range by the end of 2025, with reports indicating that the business had already turned profitable by mid-2025 [21].
| Funding Round | Date | Amount | Lead Investors | Valuation |
|---|---|---|---|---|
| Seed | April 2022 | $7.9M | Greylock Partners | Undisclosed |
| Series A | November 2024 | ~$22M | Undisclosed | Undisclosed |
| Series B | January 2025 | $105.5M | Emergence Capital, GV | ~$700M |
| Total | ~$135M |
The core workflow of Bolt.new follows a prompt-to-application pattern:
This entire cycle, from prompt to running application, typically completes in under a minute for simple projects. More complex applications with multiple pages, authentication, and database integration may take several minutes as the AI scaffolds the project in stages.
The technical backbone of Bolt.new is StackBlitz's WebContainers technology. WebContainers provide a full Node.js environment running in the browser via WebAssembly. This means that unlike many AI code generators that only produce static code, Bolt.new actually runs the generated application in real time. The system supports:
Because all code execution happens within the browser's security sandbox, there is no risk of generated code affecting the user's local machine or accessing sensitive files.
Bolt.new launched with Claude 3.5 Sonnet as its default model and has since expanded its model support significantly. As of May 2026, Claude Sonnet 4.6 is the platform default for new prompts, replacing the previously available Sonnet 4.5 and Opus 4.5 tiers, which have been retired from the model picker [31]. Users on paid plans can switch models on a per-message basis directly inside the editor, allowing different parts of the same project to be handled by different models. A typical pattern is to scaffold with Sonnet 4.6, use Haiku 4.5 for quick edits, and reserve Opus 4.6 for the hardest reasoning or refactoring tasks.
| Model | Provider | Notes |
|---|---|---|
| Claude Sonnet 4.6 | Anthropic | Current default (February 2026); agentic execution, subagent delegation, parallel tool calls, 1M token context window [31] |
| Claude Opus 4.6 | Anthropic | Reserved for the deepest reasoning tasks; supports adjustable reasoning effort, added January 2026 |
| Claude Haiku 4.5 | Anthropic | Fast, low-cost option for quick edits and minor fixes |
| Claude Sonnet 4 | Anthropic | Previous default after the October 2025 Anthropic partnership expansion |
| Claude 3.5 Sonnet | Anthropic | Original default model at launch in October 2024 |
| GPT-4o | OpenAI | Alternative model option |
| Gemini | Alternative model option |
The choice of Claude as the default model was deliberate. Eric Simons has stated that Claude 3.5 Sonnet's code generation capabilities were the catalyst that made Bolt.new possible, as earlier models consistently failed to produce working full-stack applications [2]. Successive Sonnet upgrades have repeatedly outperformed Bolt's internal benchmarks for end-to-end app generation, including a 2026 result where Sonnet 4.6 became the first Sonnet model preferred over the previous Opus generation in Bolt's own coding evaluations [31].
Sonnet 4.6 also changed how the agent behaves inside Bolt. Rather than waiting for the user to confirm each step, the model proactively plans multi-step features, dispatches subagents to parallelize work (for example, bootstrapping a database while scaffolding UI), and calls tools in parallel where possible. This shifts Bolt's interaction pattern further away from "chat then code" toward genuinely agentic execution, which StackBlitz now markets under the broader "agentic building" framing used for the Microsoft partnership [22][31]. The same release introduced first-class MCP support, letting the agent participate in large external tool ecosystems without bespoke per-tool wiring.
Bolt.new's relationship with Anthropic has evolved from a customer-vendor arrangement into a formal strategic partnership. In October 2025, StackBlitz announced an expanded collaboration that brought Claude Sonnet 4 to every Bolt user with significantly increased capacity, removing earlier rate-limit constraints that had affected heavy users during peak periods [23]. The partnership also positioned Anthropic as the headline sponsor of the company's flagship developer event, the World's Largest Hackathon, which was officially "powered by Claude."
Anthropic gained one of its largest and most visible API customers, while Bolt.new built its product around the assumption that Claude's code-generation quality would continue to improve. The companies presented a joint case study titled "The Future of Development: Anthropic x Bolt.new" at SXSW 2026 [24].
Bolt.new supports a broad range of JavaScript and TypeScript frameworks and tools, reflecting its full-stack ambitions. Because WebContainers support the full npm ecosystem, any JavaScript library available through npm can theoretically be used in a Bolt.new project.
| Framework/Technology | Type | Notes |
|---|---|---|
| React | Frontend library | Most commonly used framework on the platform |
| Next.js | Full-stack framework | Supports server-side rendering and API routes |
| Vue.js | Frontend framework | Full support including Vue 3 Composition API |
| Svelte / SvelteKit | Frontend/full-stack | Lightweight framework option |
| Astro | Static site generator | Supported for content-focused sites |
| Angular | Frontend framework | Full support for Angular applications |
| React Native (Expo) | Mobile development | Cross-platform mobile app support via official Expo integration |
| Tailwind CSS | CSS framework | Default styling solution for most generated projects |
| TypeScript | Language | Full TypeScript support across all frameworks |
| PostgreSQL | Database | Supported through Bolt Cloud built-in databases |
| Supabase | Backend-as-a-service | Integrated for authentication and database needs |
| Prisma | ORM | Database schema management and querying |
The entire development workflow takes place within the browser. Users can view and edit source code in a built-in code editor, see a live preview of the running application, access a terminal for command execution, and manage files through a file explorer panel. This approach eliminates the traditional setup friction of cloning repositories, installing dependencies, and configuring local development environments.
After generating an initial application, users can provide follow-up prompts to modify, extend, or fix the application. The AI modifies the existing codebase rather than regenerating from scratch, preserving the user's previous work. This conversational iteration loop is one of Bolt.new's core interaction patterns.
Bolt.new supports one-click deployment to production. Originally, all deployments were handled through Netlify, leveraging a partnership between StackBlitz and Netlify. In 2025, Bolt introduced its own hosting service, Bolt Cloud, which became the default for new projects. Users can still choose to deploy to Netlify if they prefer [9].
Bolt V2, introduced in October 2025, marked a significant evolution of the platform from an experimental vibe coding tool to a production-grade development environment. The centerpiece of V2 is Bolt Cloud, a hosting and backend platform that bundles several previously separate concerns into a single integrated system [10].
Bolt Cloud includes:
| Capability | Description |
|---|---|
| Built-in databases | Unlimited PostgreSQL databases created automatically based on project needs; no third-party database configuration required |
| Managed hosting | Production hosting with custom domain support and DNS management |
| Native authentication | Built-in auth system, eliminating the need for Supabase Auth, Auth0, or similar services |
| Edge functions | Server-side logic executed at the edge for low-latency responses |
| File storage | Managed file and asset storage |
| Analytics | Live user data, traffic sources, page views, bandwidth, and geographic distribution |
| Security reviews | Automatic vulnerability scanning that flags potential issues and allows one-click fixes |
| Version history | Visual browsing and one-click restoration of previous project versions |
| SEO tools | Built-in SEO management for deployed sites |
| Payment integration | Streamlined payment setup for SaaS applications |
Bolt Cloud represents a strategic shift from being purely a code generation tool to becoming a more complete application platform. By bundling hosting, databases, and authentication together, Bolt.new aims to solve what the company calls "vibe coding's fatal flaw": the gap between generating an application and actually deploying it for real users [10].
In February 2025, StackBlitz announced an official partnership with Expo, the company behind the most widely used React Native tooling and ecosystem, to bring native iOS and Android development to Bolt.new [25]. The integration allows users to describe a mobile app in natural language and receive a working React Native project, complete with Expo Router for navigation, NativeWind for styling, and a basic state management setup. Crucially, the workflow runs entirely in the browser, which means users can build iPhone apps from a Windows machine without installing Xcode or any platform-specific tooling.
The core workflow uses an Expo starter template inside Bolt.new. Once the AI scaffolds the project, users can scan a QR code with the Expo Go app on a physical phone to see the app running on real hardware in seconds. This preview loop dramatically tightens the iteration cycle compared with traditional React Native development, where setting up emulators or device builds typically takes hours.
The Expo integration carries one important limitation that StackBlitz has been transparent about. Because Bolt.new runs inside WebContainers, which cannot execute Xcode, Android Studio, or the EAS (Expo Application Services) build pipeline, the platform produces React Native source code but does not directly generate signed iOS or Android binaries [26]. Users who want to publish to the App Store or Google Play must export the code and run an EAS build separately, often by connecting their Expo account. Bolt and Expo have publicly stated that bringing EAS Build into Bolt is a roadmap item, after which a full prompt-to-store pipeline would be possible [32].
Bolt.new introduced Figma import capabilities, allowing users to convert Figma designs into functional code. The integration, powered by Anima, lets users copy a link to a Figma frame and paste it directly into Bolt's chat interface. Within seconds, the design is converted into a working application with preserved layout, spacing, and styling decisions [11]. In a 2025 update, Bolt expanded this feature to allow importing Figma frames into existing projects mid-development, not just during initial project creation. Every user receives 3 free Figma conversions per month, with additional conversions consuming tokens (50,000 to 200,000 tokens depending on design complexity).
As of early 2026, Bolt.new supports generating AI images directly from the chat interface, with features including transparent background support and WebP format conversion. This allows users to create custom graphics and assets for their applications without leaving the platform.
Bolt.new added support for Model Context Protocol (MCP) servers in early 2026, allowing users to connect external tools like Notion, Linear, and GitHub directly into their development workflow. This enables the AI agent to interact with external services and pull in data or trigger actions without requiring the user to switch between applications [12].
Bolt.new added team features in 2025 and 2026, including shared workspaces, team templates for standardized project creation, admin controls, and collaborative editing. The Teams plan allows organizations to create reusable templates that enforce consistent project structure and configuration across team members. These features target professional teams and small businesses rather than just individual developers.
Other notable features introduced over the course of 2025 and 2026 include:
Through late 2025 and into 2026, Bolt.new pursued enterprise adoption aggressively. The company introduced a dedicated Bolt for Enterprise tier with custom contracts, SSO, audit logging, and dedicated support. Procurement is available through AWS Marketplace at a reported $100,000 annual base price, with additional charges of $150 per seat and a token overage rate of $0.01 per token beyond the included allocation [27]. The enterprise push has been one of the company's main paths to durable, large-contract revenue beyond self-service Pro subscriptions.
The enterprise offering also targets a specific organizational pain point: design-system fidelity. Large companies typically spend significant engineering time translating Figma component libraries into production code. Bolt for Enterprise positions itself as a way to compress that loop by importing an organization's design system directly into reusable team templates, then letting non-engineers prompt new internal tools that automatically inherit the approved component library, theme tokens, and accessibility patterns [27].
On May 5, 2026, StackBlitz announced a major collaboration with Microsoft that listed Bolt.new in Microsoft Marketplace and integrated it with Microsoft 365 and Microsoft 365 Copilot [22][28]. The deal adds three concrete capabilities for enterprise customers:
| Capability | Description |
|---|---|
| Microsoft Marketplace procurement | Enterprise teams can buy Bolt.new through existing Microsoft vendor channels, bypassing lengthy net-new vendor approval workflows |
| Azure-native deployment | Applications generated in Bolt.new can be deployed directly into a customer's own Microsoft Azure environment, keeping infrastructure and data inside the customer's tenant |
| Microsoft 365 and Copilot integration | Bolt becomes available inside Microsoft Teams and the Microsoft 365 Copilot surface, letting employees launch agentic builds from within tools they already use daily |
Applications deployed to Azure can use Microsoft Entra for identity and access management and Microsoft Defender for automated compliance checks, addressing two common enterprise blockers for AI-generated software. In announcing the partnership, Bolt.new claimed that 75% of the Fortune 500 had adopted the platform in some capacity by mid-2026, a figure cited in StackBlitz's own enterprise materials and repeated in Microsoft's press release [22].
The Microsoft partnership stacks on top of an earlier AWS Marketplace listing, meaning Bolt.new is now formally available through two of the three largest cloud vendor marketplaces, a footprint that differentiates it from most other AI app builders, which remain primarily consumer or prosumer products.
Bolt.new uses a token-based pricing model, where usage is metered by the number of AI tokens consumed during interactions. As of 2026, unused tokens from paid plans roll over for one additional billing period, making them valid for up to two months total [13].
| Plan | Monthly Price | Token Allocation | Key Features |
|---|---|---|---|
| Free | $0 | 1M tokens/month (300K daily cap) | Unlimited databases, public/private projects, Bolt branding on deployed sites, 10MB file upload limit |
| Pro | $20 | 10M tokens | Full model access, Bolt hosting, custom domains, no Bolt branding |
| Pro 50 | $50 | 26M tokens | Higher token ceiling for moderate usage |
| Pro 100 | $100 | 55M tokens | For heavy usage |
| Pro 200 | $200 | 120M tokens | Highest individual tier |
| Teams | $30/member | Shared allocation | Admin controls, shared workspaces, team templates |
| Enterprise | Custom (AWS Marketplace base ~$100K/yr + $150/seat) | Custom | Custom terms, dedicated support, SSO, audit logging, Azure deployment |
Annual billing provides a 10% discount across all paid plans (e.g., the Teams plan drops to approximately $27 per member per month with annual billing) [13]. The free tier is sufficient for experimenting with the platform but runs out quickly for sustained development work, particularly when working on complex projects that require many iterative prompts.
Usage on Bolt.new spans a wide range of project types, but a handful of categories dominate the public showcase, the World's Largest Hackathon submissions, and StackBlitz's enterprise case studies.
| Use case | Typical builder | Notes |
|---|---|---|
| Marketing sites and landing pages | Founders, marketers | Most common first-project type; usually a single Next.js or Astro page deployed in minutes |
| Internal tools and dashboards | Operators inside larger companies | Admin panels, CRUD frontends over existing APIs, quick reporting dashboards |
| MVPs for early-stage startups | Solo founders, indie hackers | Used to validate ideas before hiring engineers; many graduates of the World's Largest Hackathon fall in this bucket |
| Internal SaaS prototypes | Product managers, designers | Lets non-engineers prototype interactions tied to real data rather than static Figma mockups |
| Mobile prototypes via Expo | Designers, content creators | iOS or Android scaffolds for early user testing, with EAS Build still required for store submission |
| Educational and demo apps | Course creators, conference speakers | Live-coded apps used in talks and tutorials, shared as project links |
| Personal and side projects | Hobbyists | Small utilities, portfolio sites, and personal automation tools |
StackBlitz has explicitly framed Bolt.new as collapsing the prototype-to-production gap. The combination of in-browser execution, one-click hosting, and built-in auth and databases is intended to make it possible for a single non-engineer to ship a working, monetizable web app, a workflow that the company highlights in case studies from solo founders who have launched paid SaaS products on Bolt alone [21]. The May 2026 Microsoft partnership extended this same framing into corporate IT environments, with internal business apps deployed to a customer's own Azure tenant as the primary enterprise use case [22].
bolt.diy is the official open-source version of Bolt.new, maintained under the StackBlitz Labs GitHub organization. Originally created by content creator Cole Medin under the name "oTToDev," the project was adopted by StackBlitz in an official partnership and renamed to bolt.diy. It allows users to run a Bolt-like AI app builder locally or self-hosted, using any LLM of their choice [14].
As of early 2026, bolt.diy has accumulated over 19,000 GitHub stars and 10,400 forks, reflecting strong community interest. The project supports over 19 LLM providers:
| Provider Category | Supported Providers |
|---|---|
| Cloud providers | OpenAI, Anthropic, Google Gemini, Groq, xAI, DeepSeek, Mistral, Cohere, Together AI, Perplexity, HuggingFace, OpenRouter, Moonshot (Kimi), Hyperbolic, GitHub Models, Amazon Bedrock |
| Local providers | Ollama, LM Studio, OpenAI-compatible endpoints |
Key features of bolt.diy beyond basic LLM flexibility include:
In early 2025, StackBlitz announced the Bolt 100K Open Source Fund, committing $100,000 to support open-source web infrastructure projects and libraries that the Bolt ecosystem relies on. The company also joined the Open Source Pledge, an initiative led by Sentry for companies to contribute financially to open-source maintainers [15]. This fund supports ecosystem tools like pkg.pr.new (continuous releases for GitHub projects) and other community-driven infrastructure.
In June 2025, Bolt.new ran what it branded the World's Largest Hackathon, a four-week global event that drew over 130,000 registered builders and a prize pool exceeding $1 million [29]. Submissions had to be built primarily inside Bolt.new and were judged across global, regional (Americas, APAC, EMEA), and themed tracks including Make More Money, Blockchain, and Silly Shit. The event was officially powered by Anthropic's Claude, reinforcing the deepening commercial alignment between the two companies. StackBlitz used the hackathon both as a marketing flywheel (with thousands of viral demos) and as a recruiting and discovery tool for emerging founders building on top of the platform.
According to StackBlitz, the hackathon helped push user counts beyond 5 million and seeded a pipeline of high-quality showcase apps used in subsequent marketing.
Bolt.new's community extends across several platforms. The official Discord server (shared with StackBlitz) serves as the primary community hub for support, feature requests, and project showcases. The platform's viral nature on social media, particularly X (formerly Twitter) and TikTok, has been a significant driver of user acquisition. Creators regularly post demonstrations of building complete applications in minutes, contributing to the broader vibe coding movement popularized by Andrej Karpathy in February 2025 [16].
Bolt.new operates in the rapidly growing AI app builder market, where several well-funded competitors are vying for dominance.
Lovable (formerly GPT Engineer) is perhaps Bolt.new's most direct competitor. Like Bolt.new, Lovable generates full-stack web applications from natural language prompts and has seen explosive growth. Lovable focuses on React and Supabase-based applications and reached $200 million ARR by November 2025, significantly outpacing Bolt.new on revenue. Lovable raised $330 million at a $6.6 billion valuation in December 2025 [17].
Industry reviewers in 2026 generally place Lovable ahead for true non-coders, while Bolt.new is positioned as the faster, more code-forward option for users comfortable iterating directly on source files [30]. A widely cited 2026 hands-on comparison concluded that Lovable produces the cleanest first-pass UI, Bolt produces the fastest deployed URL, and Cursor remains the most capable for ongoing maintenance of large codebases [30][33].
Cursor is an AI-native code editor that has become the dominant choice for professional developers who still want to write and review code directly. Unlike Bolt.new or Lovable, Cursor is a desktop IDE (a fork of VS Code) rather than a browser-based prompt-to-app builder. It targets a different layer of the workflow: accelerating an experienced developer's day-to-day coding rather than collapsing the entire build process into a single prompt. Many teams use Cursor and Bolt.new together, with Bolt.new producing initial scaffolding and Cursor handling deeper edits, refactors, and integration into larger existing codebases [30].
v0 is Vercel's AI-powered development tool. Originally focused specifically on frontend UI generation using React and Tailwind CSS, v0 has since expanded to offer fuller-stack capabilities with built-in databases. v0 benefits from deep integration with the Vercel deployment platform and the Next.js ecosystem. It is known for producing visually polished Next.js applications [18].
Replit offers an AI-powered online IDE with its Replit Agent feature, which can generate and deploy applications from natural language descriptions. Replit differentiates itself with support for a wider range of programming languages beyond JavaScript/TypeScript (including Python, Java, Go, and others) and its established developer community. Replit provides real-time multiplayer collaboration, similar to Google Docs, making it particularly strong for team use cases.
| Feature | Bolt.new | Lovable | v0 | Replit | Cursor |
|---|---|---|---|---|---|
| Full-stack generation | Yes | Yes | Expanding (originally frontend) | Yes | No (developer IDE) |
| In-browser execution | Yes (WebContainers) | No (cloud-based) | No (cloud-based) | Yes (cloud VM) | No (desktop) |
| Code runs locally | Yes (on user's device) | No | No | No | Yes |
| Default AI model | Claude Sonnet 4.6 | Claude | Custom v0 models | Proprietary | User choice |
| Built-in hosting | Yes (Bolt Cloud) | Yes (Lovable Cloud) | Yes (Vercel) | Yes | No |
| Built-in database | Yes (Bolt Cloud) | Supabase integration | Yes (built-in) | Yes (built-in) | No |
| Native authentication | Yes (Bolt V2) | Via Supabase | No | Yes | No |
| Mobile app support | React Native (Expo) | Limited | No | Yes | Via plugins |
| Language support | JavaScript/TypeScript | JavaScript/TypeScript | JavaScript/TypeScript | Multi-language | Multi-language |
| Real-time collaboration | Limited | Multiplayer workspaces | Individual focus | Real-time multiplayer | Limited |
| Figma import | Yes | No | No | No | No |
| Open-source version | bolt.diy (community) | No | No | No | No |
| Starting price | $20/month | $20/month | $0 (credit-based) | $0 (free tier) | $20/month |
| Estimated ARR (mid-2026) | ~$80-100M | ~$250M+ | Undisclosed | Undisclosed | $500M+ |
Beyond the major players, additional competitors have entered the AI app builder space:
Bolt.new's revenue trajectory is among the most dramatic in SaaS history, making it the second fastest-growing software product ever behind ChatGPT.
| Milestone | Timeframe | Notes |
|---|---|---|
| $60K ARR | Day 1 (October 4, 2024) | First-day revenue from launch |
| $80K ARR | Day 2 (October 5, 2024) | Rapid acceleration |
| $4M ARR | ~30 days (November 2024) | Initial viral growth phase |
| $20M ARR | ~60 days (December 2024) | Continued acceleration |
| $40M ARR | ~5 months (March 2025) | With only 35 employees |
| 1M+ websites deployed | March 2025 | In partnership with Netlify |
| 3M registered users | March 2025 | 1M monthly active users |
| 5M+ total users | Mid-2025 | Adding 1M+ users per month |
| 130,000+ hackathon builders | June 2025 | World's Largest Hackathon, $1M+ prize pool |
| Reported profitability | Mid-2025 | According to Sacra equity research |
| 7M+ total users | December 2025 | About 14 months after launch |
| $80-100M ARR (analyst projection) | End of 2025 | Per Sacra and other equity research firms |
| 75% of Fortune 500 adoption | May 2026 | Cited in Microsoft Marketplace announcement |
The revenue efficiency is notable: reaching $40 million ARR with a team of approximately 35 people translates to over $1 million in ARR per employee, a figure that surpasses most SaaS companies at similar stages [19]. The $700 million valuation at Series B implied a 17.5x revenue multiple based on the $40 million ARR run rate. By mid-2025, gross margins were reported in the 40% range, reflecting the heavy underlying cost of inference passed through from Anthropic [21].
Third-party reviewers in 2026 broadly agree on the same picture: Bolt.new is the fastest route from prompt to a working, deployed URL, but the code it produces typically requires more debugging than the equivalent output from Lovable or Cursor [30][33]. A March 2026 hands-on comparison reported that Lovable produced cleaner first-pass UI, Cursor handled multi-module applications with hundreds of files more reliably, and Bolt produced the most bugs in non-trivial projects but reached a live demo first [33]. For hackathons, demos, and early prototypes, several reviewers described Bolt as effectively unbeatable, while recommending that maintenance-heavy projects graduate to a proper development environment or be rebuilt in Cursor once a feature set stabilises.
Within the developer community, Bolt.new has accumulated a reputation as the strongest public showcase of what WebContainers can do, with the Anthropic alliance treated as a signal that Bolt is one of the most demanding production users of frontier code models. Critics, particularly senior engineers, have been more skeptical, pointing to the gap between the marketing of "production-grade" apps and the realities of operating AI-generated codebases at scale. Reviews aimed at non-engineers tend to be more enthusiastic, framing Bolt's bugginess as a tolerable trade-off for the ability to ship a usable product without writing code.
Despite its impressive growth, Bolt.new has faced several criticisms from users and the developer community.
Code quality concerns: Generated applications sometimes contain inefficient code patterns, duplicated logic, or suboptimal architectural decisions. Professional developers have noted that the AI can "lose context" on larger projects, leading to inconsistencies as the codebase grows [20]. Independent reviewers in 2026 found that Bolt, Lovable, and similar prompt-to-app tools all tend to hit a wall around 15 to 20 components, after which the AI begins to drop context, reintroduce earlier bugs, and require manual intervention to keep the project coherent [33].
Token consumption: Heavy users frequently report that token allocations run out faster than expected, particularly when working on complex applications that require many iterations. A single complex project can easily exhaust the Pro plan's 10 million tokens in a few days of active development.
Debugging difficulty: When the AI generates code that does not work as expected, diagnosing and fixing the issue can be challenging for non-technical users, which is ironic given that the platform's primary appeal is accessibility to non-developers.
Backend limitations: While Bolt.new supports full-stack development in principle, its strength lies more in frontend and simple backend logic. Complex backend architectures, microservices, and sophisticated database schemas remain challenging for the AI to generate reliably.
Framework flexibility vs. depth: While Bolt.new supports many frameworks, the quality of generated code can vary significantly between frameworks. React projects tend to produce the best results since the underlying models have been trained on more React code. Less popular frameworks may produce more errors.
Scalability of generated apps: Applications built with Bolt.new are well-suited for prototypes, MVPs, and small-to-medium-scale projects. However, large-scale production applications with complex state management, performance optimization needs, and enterprise-grade security requirements typically require significant manual refinement by experienced developers.
Mobile binary gap: The Expo integration produces React Native source code but cannot directly build signed iOS or Android binaries from inside the browser, leaving a meaningful gap between Bolt's in-browser preview and an actual App Store submission [26].
Margin compression: With gross margins reported around 40% in 2025, Bolt.new operates on thinner unit economics than typical SaaS businesses because most of each user dollar passes through to model providers. Pricing changes by Anthropic, OpenAI, or Google could materially affect profitability.
Bolt.new has been closely associated with the rise of vibe coding, a term coined by Andrej Karpathy on February 2, 2025, to describe a style of programming where the developer "fully gives in to the vibes, embraces exponentials, and forgets that the code even exists" [16]. Collins Dictionary selected "vibe coding" as its Word of the Year for 2025, reflecting the cultural impact of tools like Bolt.new that make software creation accessible to non-programmers.
The platform has been particularly popular among designers, entrepreneurs, and non-technical founders who want to build functional prototypes or MVPs without hiring developers. The combination of natural language input, instant visual feedback, and one-click deployment has lowered the barrier to software creation in ways that were not possible before AI-powered code generation.
As of May 2026, Bolt.new is one of the leading AI app builders and has shifted noticeably toward enterprise distribution. Recent updates have focused on model upgrades, partnership-driven distribution, and infrastructure expansion. Key developments in 2026 include:
The competitive landscape has intensified significantly since Bolt.new's launch. Lovable has outpaced Bolt.new in revenue terms, reaching reported ARR figures well above Bolt.new's trajectory. Cursor, in a different segment, has crossed $500 million in ARR and now operates at a different order of magnitude as a developer-IDE business. However, Bolt.new retains a distinctive technical advantage through its WebContainers technology, which provides true in-browser code execution rather than relying on cloud-based compute. This approach offers faster feedback loops, offline capability, and a development experience that more closely resembles working in a local IDE. The Microsoft partnership, taken together with the existing AWS Marketplace listing and the Anthropic alliance, has positioned Bolt.new as one of the few AI app builders with deep distribution into enterprise IT.
Bolt.new, along with its competitors, is riding a fundamental shift in how software is created, one that is blurring the line between technical and non-technical builders and redefining what it means to be a software developer.