South Korean startup WEVEN has launched the professional version of its AI web production platform, Zaemit.ai, adding conversational back-end implementation, a team workspace, and a reusable block system—all while promising deeper integration with Windows developers’ primary tool, Visual Studio Code. The July 2026 release targets agencies, designers, and freelancers who juggle multiple client sites, moving beyond simple AI page generation to a collaborative workflow that stays involved long after the initial build.

What actually changed: workspaces, blocks, and AI that writes back-end code

WEVEN bills Zaemit.ai as a multi-agent service where specialized AI agents handle planning, design, coding, and back-end tasks. The latest update introduces three pillars designed for professional teams: a Workspace system, Block Management, and a conversation-driven back-end implementation tool.

The Workspace acts as a central command center. From a single dashboard, users can see project status, invite team members, manage client communication, and track domain registrations and hosting expiration dates. For an agency handling 20 client sites, that means no more scrambling over missed renewal emails—the dead dates are right on screen.

Block Management lets teams save frequently used UI components—hero sections, navigation bars, footer layouts—and attach detailed AI generation guidelines to each block. When building a new page for a repeat client, you can drop in a saved block and the AI will honor its typography, spacing, and color rules, keeping the brand identity consistent without manual rework. WEVEN says this cuts errors on repetitive sub-pages and speeds up template-based projects.

The most ambitious addition is an AI agent that implements back-end functionality through chat. In demonstrations, asking “Add a bulletin board” or “Put in a reservation feature” triggers the creation of application logic and data handling. This isn’t just front-end scaffolding; the platform claims it can generate working back-end infrastructure. WEVEN describes it as part of a “Full-Lifecycle Platform” strategy, spanning from site generation to images, code, and operation.

Under the hood, the update debuts a proprietary design model called Apollon. Evolved from WEVEN’s earlier Maker 3.0/WebGen.Agent technology, Apollon has been fine-tuned on web-industry classification data and layout systems. The company says it produces entire sites with consistent fonts, margins, colors, and components without users repeatedly specifying style rules. Apollon also handles image generation in context, aiming for visual harmony across all pages.

WEVEN claims Apollon reduces token consumption and generation costs compared with general-purpose large language models, though no specific benchmarks were provided. Prospective users should test those savings against their own usage patterns—vendor efficiency claims rarely translate one-to-one in real workflows.

What it means for you, depending on who you are

For a freelancer building cafe homepages or boutique portfolio sites, the block system is a straightforward time-saver. Save your go-to layout, reuse it across clients, and let the AI tweak content while preserving your hard-won design rules. The workspace dashboard means you won’t forget which client’s domain is about to lapse—a surprisingly common headache for solo operators.

For a small agency, the collaborative features become more critical. Invite team members to a project, check real-time status, and communicate inside the tool. The back-end agent could let designers or project managers spin up functional prototypes—say, a reservations feature for a restaurant client—without pulling a developer in for the first iteration. That potentially shortens client feedback loops and reduces billable hours on routine tasks.

For Windows developers, the most immediately useful piece is the existing Visual Studio Code extension. WEVEN supports Model Context Protocol (MCP) integration, so you can call Zaemit from Claude or ChatGPT within VS Code. You can also import HTML created by external AI tools into the Zaemit editor for refinement and preparation for hosting. That makes the platform a possible sidekick in a developer’s existing workflow, rather than a browser-only silo. However, the generated back-end code is a black box: you won’t know exactly how authentication is handled, what database permissions are set, or whether the code follows security best practices until you review it. WEVEN says it has built infrastructure to manage input data, but details on data protection and auditability remain thin. For production sites that touch real customer data, treat the back-end output the way you would any auto-generated code: inspect, test, and harden.

For IT pros evaluating the tool for their organization, the key question is ownership. Who holds the rights to the generated code? What happens to your data when the AI agent processes a “reservation system” request? WEVEN hasn’t published a detailed data processing agreement or an open-sourcing roadmap for the generated code. Until those documents appear, Zaemit.ai is best treated as an internal prototyping accelerator, not a replacement for traditional development pipelines.

How we got here: from simple AI pages to a professional ecosystem

WEVEN was founded in 2024 and quickly gained traction in South Korea, reaching 3 billion won in cumulative revenue and securing 4 registered patents related to AI web technology, with 7 more PCT patents pending. The initial Zaemit service launched last October, focusing on letting anyone produce a website from natural language descriptions. By April 2026, it had topped 30,000 monthly active users and 300,000 cumulative users.

The company’s early models—Maker 3.0 and WebGen.Agent—could assemble pages but lacked the workflow features that agencies demand. Zaemit.ai fills that gap by layering project management, reusable components, and back-end generation on top of the core AI. The VS Code extension arrived earlier, extending the tool’s reach to developers who prefer working in a code editor. With the July update, those extension capabilities are now natively integrated into the web editor as well, creating a more fluid experience across interfaces.

WEVEN’s roadmap points to additional agents for deployment and operation, suggesting the company wants to own more of the web production lifecycle. CEO Junghwan Kim called this update “a turning point that helps professionals collaborate and streamline repetitive tasks.” The stated goal: a platform where design perfection and feature implementation come from a single conversation.

What to do now

If Zaemit.ai sounds relevant to your work, begin with a non-production project. Sign up for the free tier (if available) or a trial and build a test site that mimics a real client scenario: include a contact form, a user authentication flow, and a simple back-end feature like a blog comment system. Then:

  • Export the generated code and review it closely—especially the back-end parts. Check for hardcoded credentials, SQL injection vulnerabilities, and missing input validation.
  • Test the workspace with a small team. How intuitive is the client communication and domain tracking? Does the dashboard feel responsive, or does it add friction?
  • Push the block management system by creating a brand template and having a colleague build a new sub-page. Is the AI guidance truly consistent, or do you still need to fix alignment and color issues?
  • If the VS Code extension matters to you, test MCP integration. Import some HTML from another AI tool into Zaemit and see how the editor handles it. Does it preserve your code structure, or does it attempt to reformat everything?
  • Finally, gauge cost. Track how many tokens a typical project consumes and compare that with your current workflow’s resource usage. WEVEN’s cost-saving claims must be validated against your real project mix.

For teams that handle sensitive client data—medical records, financial information, personal identifiers—hold off on using the back-end agent for anything production-facing until WEVEN publishes a clear security whitepaper and data processing terms. The platform’s conversational implementation might generate a working reservation system, but if it stores customer information in a default configuration that lacks encryption at rest, you’ve created a liability.

What to watch next

WEVEN plans a global push following this professional release, which means more users, more integrations, and likely more scrutiny of its back-end generation claims. Keep an eye on two things: first, whether third-party researchers or early adopters report on the actual quality and security of the AI-generated back-end code. Second, watch for the announced import feature that will let you bring external AI-generated HTML into Zaemit for editing—if that works smoothly, the platform could become a convenient bridge between different AI coding tools and a hostable site.

The bigger story is whether an AI web builder can truly own the full lifecycle. If WEVEN ships agents for deployment and operation as promised, Zaemit could evolve from a design tool into a genuine production alternative for certain classes of websites. For now, it’s a promising glimpse of what happens when AI meets the unglamorous work of managing client sites—renewals, revisions, and reusable components—rather than just generating another shiny landing page.