WeWeb MCP

Product Introduction

1. Definition: WeWeb MCP (Model Context Protocol) is a specialized server that integrates the WeWeb visual development platform with the MCP ecosystem. Technically, it is a remote MCP server that exposes a suite of tools, allowing AI agents to read from and write to a live WeWeb project's structure, including its pages, components, workflows, data models, and integrations.
2. Core Value Proposition: It exists to bridge the gap between AI-assisted code generation and visual, production-ready application development. Its primary value is enabling developers and product teams to leverage the reasoning and speed of Large Language Models (LLMs) like Claude, GPT, or Gemini within a controlled, reviewable, and maintainable visual editor, using their own AI credits and model subscriptions.

Main Features

1. Agent-Agnostic MCP Integration: The product functions as a standards-compliant MCP server. It works by accepting structured tool calls from any MCP-compatible client (e.g., Claude Desktop, Cursor, Windsurf). The agent sends commands via the protocol, and the WeWeb MCP server executes them directly against the user's authorized WeWeb project via a secure API. This decouples the AI model from the build platform.
2. Granular Project Scope Control: Users can configure their MCP client to allow or restrict specific tools. This enables precise control over what the AI agent can modify. For example, an agent can be scoped to work only on a specific page's components, a single backend workflow, or the design system, preventing unintended changes across the entire application. This is managed through the MCP client's configuration file, not within WeWeb itself.
3. Bi-Directional Visual Editor Synchronization: Every change an AI agent makes through the MCP tools is instantly reflected in the user's WeWeb visual editor. This is not a one-time code generation; it's a live connection. The "visual diff" is inherent, as the editor displays the current state of the project. Users can then continue iterating with the AI agent or switch to manual no-code/low-code editing within WeWeb, with all changes persisted to the same project file.

Problems Solved

1. Pain Point: The "black box" problem of AI-generated code, where outputs are static codebases that are difficult to review, edit, and maintain visually, especially for cross-functional teams. It also solves the vendor lock-in of proprietary AI credits within development platforms.
2. Target Audience: The primary personas are Full-Stack Developers and Technical Product Managers who use AI coding agents (Claude Code, Cursor) and need to ship production applications faster; Low-Code/No-Code Developers in enterprises who require governance and review capabilities for AI-assisted builds; and Agency Teams that need to rapidly prototype client applications from PRDs and designs while maintaining a handoff-ready, editable project.
3. Use Cases: Rapid Prototyping from PRDs: Transforming a product requirements document into a structured WeWeb app with pages, data models, and workflows. Design System Implementation: Converting Figma variables or a design.md file into a consistent set of reusable visual components in WeWeb. Legacy App Refactoring: Using an AI agent to analyze an existing WeWeb project to standardize naming conventions, remove unused variables, or restructure complex workflows. Context-Aware Frontend Building: Pairing the WeWeb MCP with a backend MCP server (e.g., for Supabase or Xano) so the AI agent can build the frontend interface with full knowledge of the existing data schema and API endpoints.

Unique Advantages

1. Differentiation: Unlike AI features baked into platforms (which use proprietary models and credits), WeWeb MCP is model-agnostic and uses the user's own AI tokens. Unlike standalone AI code generators that output raw code, its output is a live, visually editable project in a mature low-code platform. Compared to traditional no-code tools, it introduces agentic automation without sacrificing editorial control.
2. Key Innovation: The core innovation is treating the visual application project as a malleable data structure accessible via a standardized protocol (MCP). This allows external AI agents to function as an intelligent editing layer on top of WeWeb's core engine. The "agent-in-the-loop" workflow, where every AI change is a commit to a visual canvas, represents a novel paradigm for human-AI collaboration in software development.

Frequently Asked Questions (FAQ)

1. How does WeWeb MCP handle authentication and project security? The MCP server requires you to sign into your WeWeb account and authorize access, creating a secure, scoped connection. The agent only has access to the projects and tools you've permitted through your MCP client configuration, and all operations are performed under your user account within WeWeb's permission system.
2. Can I use WeWeb MCP to build backend logic and database structures, or is it only for frontend UI? Yes, the MCP tools provide capabilities for both frontend and backend development. An agent can create and modify data tables (collections), define fields and relationships, set up authentication rules, configure serverless workflows, and manage API connections, in addition to building pages and components.
3. What happens if the AI agent makes an error or a change I don't want? Since all changes are made in real-time to your WeWeb project, you can immediately see the result in the visual editor. WeWeb has built-in version history and undo/redo functionality, allowing you to revert any specific change. You can then refine your prompt or adjust the MCP tool permissions to guide the agent more effectively.
4. Is the code generated by my AI agent through WeWeb MCP exportable? Yes, one of WeWeb's core features is the ability to export the entire project as clean, standard Vue.js 3 code. Any application built or modified with the assistance of an AI agent via MCP retains this capability, allowing for full self-hosting or further custom development in a traditional IDE.
5. Does using WeWeb MCP require knowledge of the MCP protocol or server configuration? Basic setup requires pasting a provided JSON configuration into your MCP client (like Claude Desktop), which is a one-time action. No deep protocol knowledge is needed. However, advanced users who wish to customize tool access or build complex multi-server MCP workflows will benefit from understanding MCP concepts.