Lovable (AI app builder)
Last reviewed
Apr 28, 2026
Sources
32 citations
Review status
Source-backed
Revision
v6 ยท 4,968 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Apr 28, 2026
Sources
32 citations
Review status
Source-backed
Revision
v6 ยท 4,968 words
Add missing citations, update stale details, or suggest a clearer explanation.
Lovable is an AI-powered application builder that generates full-stack web applications from natural language descriptions. Originally launched as GPT Engineer in mid-2023, the platform was rebranded to Lovable in December 2024 to signal a shift from developer-focused tooling to a broader audience of non-technical builders [1]. Founded by Anton Osika and Fabian Hedin in Stockholm, Sweden, Lovable has become one of the fastest-growing technology companies in Europe, reaching $200 million in annualized recurring revenue within eight months of rebranding and a $6.6 billion valuation by December 2025 [2][3].
Lovable generates applications using React, Supabase, and Tailwind CSS, deploying them as production-ready web applications with authentication, databases, and real-time capabilities built in. The platform competes directly with Bolt.new, v0 by Vercel, and Replit in the AI app builder market, and its rapid growth has made it a poster child for the vibe coding movement that gained momentum in 2025.
In June 2023, Anton Osika, then serving as CTO of the Stockholm-based e-commerce startup Depict.ai, created an open-source project called GPT Engineer. The tool demonstrated how large language models could write functional code from simple natural language prompts. GPT Engineer used OpenAI's GPT models to generate entire codebases from text descriptions, and the project struck a nerve with the developer community [4].
The repository went viral on GitHub, accumulating over 52,000 stars and becoming one of the fastest-growing GitHub repositories of 2023. A waitlist of 27,000 users formed by late 2024. The explosive interest validated Osika's thesis that AI-generated software was approaching a tipping point. He left Depict.ai in late 2023 to pursue the idea full-time, teaming up with Fabian Hedin to form the company initially known as GPT Engineer AB in November 2023 [4][5].
The transition from open-source tool to commercial product involved building a hosted platform, GPT Engineer App, that wrapped the code generation capabilities in a web-based interface with deployment, collaboration, and iteration features. Unlike the open-source version, which required technical knowledge to set up and run, the commercial product was designed for users with little or no coding experience.
The commercial platform launched in 2024 and began generating revenue. Within its first four weeks, the product reached $4 million in annualized recurring revenue. By the end of its second month, that figure had climbed to $10 million ARR [6]. The speed of early adoption signaled strong product-market fit and set the stage for the rapid scaling that followed.
In December 2024, the company rebranded from GPT Engineer to Lovable. The name change reflected several strategic considerations. "GPT Engineer" tied the product to a specific model family (OpenAI's GPT) at a time when the company was expanding to support multiple AI providers including Anthropic's Claude and Google's Gemini. The name also implied a developer-centric tool, while the company's ambition was to reach a much broader audience of entrepreneurs, designers, product managers, and non-technical builders who wanted to create software without writing code [1].
The rebrand coincided with a significant acceleration in product development and growth. The new name, Lovable, conveyed the company's mission of enabling anyone to build software they love.
Lovable's core workflow is straightforward: the user describes the application they want to build in plain English (or other supported languages), and the AI generates a complete, deployable web application. The generated applications are built on a modern technology stack.
| Component | Technology | Purpose |
|---|---|---|
| Frontend | React + Vite | User interface components and interactivity |
| Styling | Tailwind CSS | Responsive design and visual styling |
| Backend/Database | Supabase | PostgreSQL database, authentication, real-time subscriptions, file storage |
| Payments | Stripe | Payment processing integration |
| Deployment | Lovable Cloud / Custom domains | Production hosting with SSL |
| Version control | GitHub integration | Code export and collaboration |
Users interact with Lovable through a chat-based interface, providing prompts that describe features, layouts, behaviors, or modifications. The AI generates or updates the codebase in response, and users can see the running application in a live preview panel.
Lovable employs a multi-model orchestration system rather than relying on a single large language model. The platform uses a smart routing approach that selects different models for different tasks based on speed, cost, and quality requirements. The system integrates models from OpenAI, Anthropic (Claude), Google (Gemini), Groq, and Cohere [7]. For fast initial processing, the platform may use smaller models like GPT-4 Mini, while more complex code generation tasks are handled by models such as Claude Sonnet. Gemini Flash serves as a default for many routine operations. This multi-provider approach also provides resilience; if one provider experiences downtime, the system can route requests to alternatives.
A key aspect of Lovable's workflow is iterative refinement. After the initial generation, users can provide follow-up prompts to add features, fix issues, change designs, or restructure the application. The platform maintains context across the conversation, so each successive prompt builds on the existing codebase rather than starting over.
This conversational approach to software development is central to the vibe coding concept, a term coined by Andrej Karpathy in February 2025 to describe a coding approach where users "fully give in to the vibes" and let AI handle the implementation details [8].
Lovable's deep integration with Supabase is one of its defining technical characteristics. Supabase provides the backend infrastructure for Lovable-generated applications, including:
With the introduction of Lovable Cloud in late 2025, the platform began offering Supabase capabilities natively, meaning users no longer need to set up a separate Supabase account. Lovable Cloud utilizes Supabase's open-source foundation to provide a production-ready environment from day one, without any manual setup steps [9]. However, users who prefer to manage their own Supabase instance can still connect an external account.
Lovable integrates with GitHub, allowing users to export their generated code to a GitHub repository. This is important for users who want to continue development in a traditional code editor, collaborate with developers, or maintain version control outside of the Lovable platform. The integration supports bidirectional sync, so changes made in GitHub can be pulled back into Lovable, and changes made in Lovable are pushed to the repository.
In late 2025 and into 2026, Lovable added support for [[Model Context Protocol]] (MCP) servers as chat connectors. The connectors let the Lovable agent pull live context from a team's existing planning, documentation, and automation tools, then turn that context into working code without forcing users to copy and paste specs. The initial set of supported MCP integrations includes:
| Tool | What Lovable can read or write |
|---|---|
| [[Notion]] | Product requirement documents, design specs, technical notes, and embedded Lovable preview blocks |
| [[Linear]] | Issues, project plans, and product tickets used to scope new features |
| [[Jira]] | Backlog items, epics, and sprint tickets from Atlassian environments |
| [[Confluence]] | Knowledge base pages, with Lovable available as a built-in partner agent that can turn a Confluence page into a working app |
| [[n8n]] | Workflow automations, which in turn integrate with more than 400 other tools including [[Salesforce]], [[Slack]], and Google Sheets |
The MCP connectors are part of Lovable's push into enterprise workflows, where product managers can move from a Linear issue or a Notion PRD to a clickable prototype without leaving the planning surface they already use [27][28].
Agent Mode launched in beta on June 30, 2025 and became the default Lovable experience for all users on July 23, 2025, the same week the company crossed [[100 million ARR]]. Agent Mode is an autonomous, multi-step reasoning loop that lets the AI think, plan, and execute work with significantly less hand-holding from the user. Rather than responding to a single prompt with a single code change, the agent can explore the codebase, read files to understand existing structure, query the [[Supabase]] database, inspect runtime logs and network requests to diagnose errors, search the web in real time for documentation, and decide on its own which subtasks to perform [25].
Lovable reports that Agent Mode reduces build error rates by up to 91 percent compared to the previous single-turn workflow, while also cutting unintended changes to working code. The mode is the foundation for the conversational planning experience inside [[Chat Mode]] and underpins the platform's multi-step debugging features [25][26].
Lovable 2.0, released in April 2025 with a major update in February 2026, introduced two distinct interaction modes [10]:
This dual-mode approach addresses a common complaint about earlier versions: that the AI would sometimes make unwanted changes when the user was only trying to discuss a problem.
Lovable offers visual editing capabilities that allow users to modify CSS properties, layout, spacing, colors, and typography through a visual interface rather than writing code or prompts. This feature targets designers and non-technical users who think in terms of visual properties rather than code syntax [10].
One of the most significant additions in Lovable 2.0 was real-time multi-user collaboration for up to 20 users per workspace. Earlier versions of the platform were limited to single-player use, which restricted its utility for teams. The collaboration feature allows multiple users to work on the same project simultaneously, viewing each other's changes in real time. Role-based access controls let workspace owners and admins manage users, while editors can modify existing projects [10].
Lovable 2.0 introduced automated vulnerability scanning that runs before deployment. When a user publishes their application, the platform scans the codebase for common security issues (particularly for projects connected to Supabase), flagging potential problems before the application goes live [10].
The platform added built-in domain purchasing, allowing users to buy and configure custom domains directly within Lovable rather than managing DNS settings through a third-party registrar. Over 10,000 domains have been connected through this feature since its introduction [10].
Introduced alongside the broader 2.0 update, Versioning 2.0 provides bookmarking capabilities, easier restores from previous states, and an improved history view. This gives users more confidence to experiment, knowing they can roll back to a known-good state at any time [10].
Anton Osika, co-founder and CEO of Lovable, was born in Sweden and studied at the KTH Royal Institute of Technology. His career began with work on particle physics at CERN, after which he became the first employee and engineer at Sana Labs, an AI-powered learning platform that went on to raise over $80 million [4][11].
In 2020, Osika co-founded Depict.ai alongside Oliver Edholm, a startup building AI-powered product recommendations for e-commerce. Depict.ai was accepted into Y Combinator's Summer 2020 (S20) batch and went on to raise $20 million from investors including Initialized Capital (Garry Tan's fund), EQT Ventures, and NorthZone. As CTO, Osika helped scale the platform to serve billions of product recommendations [4][11][12]. He left Depict.ai in late 2023 to work on GPT Engineer full-time.
Osika's decision to stay headquartered in Stockholm rather than relocate to Silicon Valley has been a notable aspect of Lovable's story. He has credited the decision with helping the company maintain focus and avoid the distractions of the Bay Area startup scene, while still being able to attract top engineering talent from across Europe [3].
Following Lovable's $6.6 billion valuation in December 2025, Osika, at 26 years old, was reported to be Europe's youngest self-made billionaire, with a net worth of approximately $1.6 billion based on his estimated 24% ownership stake [13].
Fabian Hedin is the co-founder and CTO of Lovable. Before Lovable, Hedin served as front-end engineering lead at Depict.ai and built the proptech firm TenFAST. His engineering background complemented Osika's technical leadership, and together they transformed the GPT Engineer open-source project into a commercial platform. Like Osika, Hedin holds an estimated 24% stake in the company, putting his net worth at approximately $1.6 billion following the Series B [13].
Lovable's fundraising history reflects one of the fastest valuation trajectories in European startup history.
| Round | Date | Amount | Valuation | Lead Investors |
|---|---|---|---|---|
| Pre-seed/Seed | October 2024 | $7.5M | Undisclosed | Hummingbird Ventures, byFounders |
| Pre-Series A | February 2025 | $15M | Undisclosed | Creandum |
| Series A | July 2025 | $200M | $1.8B | Accel |
| Series B | December 2025 | $330M | $6.6B | CapitalG, Menlo Ventures |
The seed round in October 2024 included notable angel investors such as Charlie Songhurst (former Microsoft executive), Adam D'Angelo (CEO of Quora), Thomas Wolf (co-founder of Hugging Face), and several DeepMind alumni [6].
The Series A round in July 2025, led by Accel at a $1.8 billion valuation, made Lovable one of the fastest companies in European history to reach unicorn status, achieving the milestone approximately two years after founding [14]. The Series B, just five months later, tripled the valuation to $6.6 billion. Participants in the Series B included NVentures (NVIDIA's venture arm), Salesforce Ventures, Databricks Ventures, T.Capital (Deutsche Telekom), Atlassian Ventures, HubSpot Ventures, Khosla Ventures, DST Global, EQT Growth, and Kinship Ventures, alongside returning investors Accel, Creandum, and Evantic [2].
Total funding raised stands at approximately $553 million across all rounds [2].
It is worth noting that while Lovable itself is not a Y Combinator portfolio company, the founders' Y Combinator experience with Depict.ai (S20 batch) gave them access to YC's network of investors and advisors, which played a role in the company's early fundraising [12].
Lovable uses a credit-based pricing model. Each prompt or action consumes a variable number of credits depending on complexity; a small request might use a fraction of a credit, while generating an entire page with multiple components will consume significantly more.
| Plan | Monthly Price | Credits | Key Features |
|---|---|---|---|
| Free | $0 | 5 daily (up to 30/month) | Private projects, unlimited workspace members |
| Pro | From $25 | 100-10,000/month + 5 daily bonus | Custom domains, remove branding badge, code mode, credit rollover |
| Business | From $50 | 100-10,000/month + 5 daily bonus | SSO, data training opt-out, reusable design templates, user credit spend limits |
Pro plan pricing scales with credit volume: 100 credits for $25/month, up to 10,000 credits for $2,250/month. Annual billing provides a 16% discount. Business plan pricing follows the same credit tiers at roughly double the Pro rates, reflecting the additional enterprise features [15].
Credit top-ups are available in 50-credit increments at $15 (Pro) or $30 (Business) per increment, with a maximum purchase of 1,000 credits at a time. Unused credits roll over month to month and expire upon plan cancellation [15].
Lovable Cloud hosting is billed separately from the subscription plan and scales with usage (requests, storage, bandwidth). Every workspace receives $25 in free monthly Lovable Cloud usage, a promotional offering available through early 2026 [9].
Lovable's growth trajectory has been exceptional, even by the standards of the AI boom. The company has been called the fastest-growing software startup in history by multiple industry analysts.
| Metric | Value | Date |
|---|---|---|
| ARR | $4M | First 4 weeks (early 2025) |
| ARR | $10M | First 2 months |
| ARR | $17M | February 2025 |
| ARR | $50M | May 2025 |
| ARR | $100M | July 2025 |
| ARR | $200M | November 2025 |
| ARR | $250M | December 2025 |
| ARR | $300M | January 2026 |
| ARR | $400M | February 2026 |
| Registered users | ~8M | November 2025 |
| Registered users | 15M+ | Early 2026 |
| Paid customers | 45,000+ | March 2025 |
| Daily new projects | 200,000+ | Early 2026 |
| Total projects built | 25M+ | First year |
| Visits to Lovable-built sites | 500M+ | First six months |
| Employees | ~146 | March 2026 |
The company crossed $100 million ARR in July 2025, just eight months after rebranding [16]. By November 2025, ARR had doubled to $200 million, and the company was approaching 8 million registered users [3][17]. In March 2026, Lovable reported adding $100 million in revenue in a single month, bringing ARR to approximately $400 million with just 146 employees. This translates to roughly $2.7 million in ARR per employee, nearly seven times the industry benchmark of $200,000 to $400,000 per employee for established SaaS companies [18].
Research firm Gartner has predicted that a new wave of unicorns will emerge by 2030 with $2 million ARR per employee. Lovable has already surpassed that threshold. The company has announced plans to grow its headcount, with approximately 70 open positions as of March 2026 [18].
Through late 2025 and into 2026, Lovable shifted from a primarily prosumer audience toward enterprise sales. Public reference customers include [[Klarna]], [[Uber]], and [[Zendesk]], all of which use the platform for internal tooling, prototyping, and customer-facing experiments. Klarna CEO Sebastian Siemiatkowski participated as an angel investor in the [[Series A]] round, alongside [[Slack]] co-founder [[Stewart Butterfield]], strengthening the link between Lovable and the European fintech and SaaS founder community [29].
To support enterprise demand, Lovable replaced its previous Teams plan with the Business tier in early 2026, adding [[single sign-on]], reusable templates, private projects by default, training opt-out, and per-user credit spend limits aimed at procurement and security teams. The company has also begun publishing reference cases that emphasize internal tool development, marketing landing pages, and rapid customer-facing prototypes inside Fortune 500 environments [29].
Although Stockholm remains Lovable's headquarters, the company opened two United States offices to be closer to its largest customer base. In April 2026, Lovable signed a lease for roughly 6,000 square feet at One Lincoln in Boston's Financial District, designating the location as the U.S. go-to-market hub focused on sales, partnerships, and customer success. A second office in San Francisco opened around the same time to recruit engineering and product talent on the West Coast and to be closer to model providers and investors. CEO [[Anton Osika]] has continued to emphasize that core engineering remains in Sweden [30].
In November 2025, Lovable acquired [[Molnett]], a European cloud infrastructure startup specializing in secure container deployments. Molnett's technology was folded into Lovable Cloud, giving the company end-to-end control of the path from natural language prompt to a running application served on its own infrastructure. The acquisition also brought a small team of cloud and platform engineers into Stockholm and accelerated the rollout of regional deployment options for enterprise customers with data-residency requirements [31][32].
Customer retention has been strong, with the company reporting 85%+ retention rates. Notable portfolio companies built on Lovable include ShiftNex ($1 million ARR in five months), Lumoo ($800,000 ARR in nine months), and Q Group (which generated $3 million in revenue in 48 hours) [2].
Lovable operates in the AI app builder market alongside several well-funded competitors.
Bolt.new is Lovable's most direct competitor. Built by StackBlitz, Bolt.new also generates full-stack applications from natural language prompts. Its key differentiator is StackBlitz's WebContainers technology, which executes generated code entirely within the browser rather than on cloud servers. Bolt.new reached $40 million ARR by March 2025 and raised $105.5 million at a $700 million valuation [19]. Bolt has since added its own backend and database capabilities powered by Supabase.
v0 is Vercel's AI-powered UI generation tool, which focuses on generating React, Next.js, and Tailwind CSS components using the shadcn/ui component library. Unlike Lovable and Bolt.new, v0 started as a frontend-only tool, though it has since expanded. v0 benefits from deep integration with the Vercel platform and the Next.js ecosystem but lacks a built-in backend story [20].
Replit offers an AI-powered online development environment with its Replit Agent feature. Replit supports a wider range of programming languages and targets both beginners and experienced developers. The company raised $400 million in early 2026, tripling its valuation to $9 billion. Replit reports over 50 million registered users and claims that employees from 85% of Fortune 500 companies use the platform [21].
Cursor operates in a related but distinct segment. While Lovable targets non-technical and semi-technical users who want to build applications through natural language, Cursor is an AI-enhanced code editor designed for professional developers who write and modify code directly. Cursor crossed $2 billion in ARR and reached a valuation of nearly $30 billion by late 2025, making it the largest company in the vibe coding space by valuation [22].
| Feature | Lovable | Bolt.new | v0 | Replit |
|---|---|---|---|---|
| Target audience | Non-technical to semi-technical | Broad (non-technical to developers) | Developers and designers | Developers and learners |
| Default stack | React + Supabase + Tailwind | Varies (React, Vue, Svelte, etc.) | React + Tailwind + shadcn/ui | Multi-language |
| Backend support | Full (via Supabase/Lovable Cloud) | Full (via Supabase) | Limited | Full |
| In-browser execution | No (cloud-based) | Yes (WebContainers) | No (cloud-based) | Yes (cloud VM) |
| GitHub integration | Yes (bidirectional sync) | Limited | Yes | Yes |
| Visual editing | Yes | No | No | No |
| Collaboration | Up to 20 users | Team workspaces | Team plans | Multiplayer editing |
| Starting paid price | $25/month | $15/month | Free tier available | $25/month |
| Valuation (latest) | $6.6B | ~$700M | Part of Vercel ($9.3B) | ~$9B |
Many development teams have adopted a combined approach, using Lovable for rapid prototyping and initial MVP creation, then transitioning to Cursor or a traditional code editor for production-grade refinement.
Lovable's generated applications use a consistent technology stack centered on React, Supabase, and Tailwind CSS. This opinionated approach has both advantages and limitations.
Advantages: The standardized stack means that every generated application follows the same architectural patterns, making it easier for the AI to produce consistent, high-quality code. Supabase integration provides a robust backend without requiring users to understand server-side programming. React and Tailwind are widely adopted, meaning generated code can be easily maintained by human developers.
Limitations: Users who need a different technology stack (e.g., Vue.js, Angular, or a custom backend) cannot use Lovable. The platform is designed around its chosen stack, and there is no option to swap out components. This contrasts with competitors like Bolt.new, which support multiple frontend frameworks.
In May 2025, security researcher Matt Palmer disclosed CVE-2025-48757, a vulnerability affecting Lovable-generated applications. The issue stemmed from insufficient or missing Row Level Security (RLS) policies in Supabase databases created by Lovable's AI. Because Lovable-generated applications embed a public anonymous key (anon_key) in client-side code for Supabase access, attackers could directly query the database and access data from tables that lacked proper RLS policies [23].
Palmer's analysis found that 170 out of 1,645 Lovable-generated web applications examined (10.3%) had critical security flaws, exposing 303 vulnerable endpoints. The exposed data included personally identifiable information (names, emails, phone numbers, home addresses), financial data (payment information, transaction histories, personal debt amounts), and developer credentials (API keys for services like Google Maps, Stripe, and eBay) [23].
Lovable responded by introducing the Security Scan feature in its 2.0 update, which checks for common vulnerabilities including missing RLS policies before deployment. The incident highlighted a broader challenge for AI app builders: generated code may appear functional while containing security gaps that non-technical users are unlikely to detect on their own.
Single-stack constraint: As noted above, Lovable only generates React/Supabase/Tailwind applications. Users who need other frameworks or backend technologies must look elsewhere.
Code ownership concerns: While Lovable allows GitHub export, the generated code can be tightly coupled to Lovable's patterns and Supabase's infrastructure, potentially creating lock-in for users who want to migrate to a different platform.
Complexity ceiling: Like all AI app builders, Lovable works well for standard application patterns (dashboards, CRUD apps, landing pages, authentication flows) but struggles with highly custom or complex logic. Users building sophisticated applications often hit a ceiling around 15 to 20 components, where the AI begins losing context and can make destructive changes to existing code [24].
Credit consumption: Users on lower-tier plans frequently report running out of credits, particularly during the early stages of development when many iterations are needed to refine the application.
Security of generated code: As demonstrated by CVE-2025-48757, AI-generated code may contain security vulnerabilities that non-technical users are unable to identify. While the Security Scan feature addresses some of these issues, it does not eliminate the risk entirely [23].
As of April 2026, Lovable is one of the dominant players in the AI app builder market. With an ARR approaching $400 to $500 million, a $6.6 billion post-money valuation from the December 2025 Series B, more than 15 million registered users, two newly opened United States offices in Boston and San Francisco, and a team of roughly 146 employees in Stockholm plus a growing US presence, the company has established itself as the category leader in Europe and a credible global challenger to [[Bolt.new]] and [[Replit]] [18][30].
The release of Lovable 2.0 addressed several longstanding limitations, including single-player restrictions, lack of visual editing, and the absence of direct code editing. The company has also begun expanding into enterprise sales, targeting corporate teams that want to accelerate internal tool development and prototyping. The Series B funds are earmarked for deeper integrations with productivity tools like Notion, Linear, Jira, and Miro; enhanced collaboration and governance features for enterprise adoption; and production infrastructure improvements [2][17].
Lovable's CEO Anton Osika has spoken publicly about the company's long-term vision of becoming "the last piece of software," a platform so capable that it can build any other software application. While that vision remains aspirational, the company's growth trajectory suggests that the AI app builder market is far from saturated, and Lovable is well-positioned to capture a significant share of it.