MCP Server

Open in ChatGPT

The webforJ Model Context Protocol (MCP) server plugs AI coding assistants into webforJ's documentation, APIs, design tokens, and scaffolding tools. Instead of guessing at framework conventions, the assistant asks the server and gets answers grounded in the real webforJ.

Use the plugin

Unless you know you want only the MCP server, install the webforJ AI plugin instead - it bundles this server with the matching Agent Skills in a single install.

What's an MCP?

Model Context Protocol is an open standard that lets AI assistants call external tools on demand. The webforJ MCP server implements this protocol so your assistant can:

• Look things up in the webforJ docs instead of hallucinating method names
• Scaffold new webforJ projects from official Maven archetypes
• Generate accessible DWC themes from a brand color
• Read the real styling surface of a DWC component, and validate any --dwc-* token before it lands in your CSS

AI Can Still Make Mistakes

The MCP server significantly improves accuracy, but AI assistants can still produce incorrect code in complex scenarios. Always review and test generated code before shipping.

Installation

For the full experience, install the webforJ AI plugin - it configures this server alongside the Agent Skills your assistant needs to use it well.

If you want only the MCP server (no skills), point your client at https://mcp.webforj.com/mcp:

Claude Code

claude mcp add webforj-mcp https://mcp.webforj.com/mcp -t http -s user

GitHub Copilot CLI

The recommended path on Copilot CLI is the webforJ AI plugin - it registers the MCP server for you in one step. For a raw MCP-only setup, see the per-client instructions in the webforJ AI repository.

VS Code + Copilot

Add to your VS Code settings:

"mcp": {
  "servers": {
    "webforj-mcp": {
      "url": "https://mcp.webforj.com/mcp"
    }
  }
}

Gemini CLI

Add to ~/.gemini/settings.json:

{
  "mcpServers": {
    "webforj-mcp": {
      "httpUrl": "https://mcp.webforj.com/mcp"
    }
  }
}

OpenAI Codex CLI

Add to ~/.codex/config.toml:

[mcp_servers.webforj-mcp]
url = "https://mcp.webforj.com/mcp"

Other clients

Cursor, Kiro, Goose, Junie, Antigravity, and any other MCP-over-HTTP client work too - they just use their own configuration format. See the per-client install guide for the exact snippet for each.

What the server can do

When the MCP server is connected, your AI assistant gains the following capabilities. Any of them can be triggered by a natural-language request - the assistant picks the right one automatically.

Target the right webforJ version

Before answering version-sensitive questions (anything styling or API specific), the assistant resolves which webforJ version you're on. It reads pom.xml when available and otherwise asks you. Every subsequent answer is scoped to that version.

Look things up in the webforJ knowledge base

The assistant can query the full webforJ knowledge base for answers grounded in the real framework. Results are scoped to what you ask about - an API question, a guide, a code sample, or the Kotlin DSL.

Example prompts:

"Find the webforJ Button component event handling examples"

"How do I set up routing with @Route in webforJ?"

"Show me a webforJ form validation sample"

Scaffold a new webforJ project

The assistant generates the correct Maven archetype command for a new webforJ app from your requirements (archetype, Spring integration, name, group).

Archetypes:

• hello-world - starter app with sample components
• blank - minimal project structure
• tabs - tabbed interface layout
• sidemenu - side navigation layout

Flavors:

• webforj - standard webforJ app
• webforj-spring - webforJ integrated with Spring Boot

Example prompts:

"Create a webforJ project called CustomerPortal using the sidemenu archetype"

"Generate a webforJ Spring Boot project with the tabs layout named Dashboard"

Available Archetypes

For the full list of archetypes, see the archetypes catalog.

Generate a DWC theme

From a single brand color, the assistant produces a complete DWC theme: primary, success, warning, danger, info, default, and gray palettes with automatic text contrast. Output includes the style sheet plus the @AppTheme / @StyleSheet wiring.

Example prompts:

"Generate a webforJ theme from brand color #6366f1"

"Create an accessible theme with HSL 220, 70, 50 as primary"

Style DWC components correctly

The assistant reads the real styling surface of each DWC component - CSS custom properties, shadow parts, reflected attributes, and slots - before writing any CSS. It can also enumerate every DWC tag and resolve webforJ Java class names (Button, TextField) to their DWC equivalents.

Example prompts:

"What CSS variables and parts does dwc-button expose?"

"Show me every slot available on dwc-dialog"

"Which DWC tag does the webforJ TextField class map to?"

Pair this with the styling-apps agent skill for end-to-end styling workflows.

Work with DWC design tokens

The assistant can list the authoritative catalog of --dwc-* tokens for your webforJ version - palette seeds, shades, surfaces, spacing, typography, borders - filtered by prefix or substring. It will also validate any CSS, Java, or Markdown source you give it against the real token catalog and flag unknown names with suggested corrections.

Example prompts:

"List every --dwc-space-* token"

"Validate app.css for any unknown --dwc-* tokens"

"What primary-palette shades are available?"

Validation catches typos and invented tokens before they ship as silently-failing CSS.

Writing good prompts

The MCP server is only consulted when your assistant thinks it's relevant. A few habits keep it engaged:

• Name the framework. Mention "webforJ" in the prompt so the assistant reaches for the MCP server instead of its general Java knowledge.
• Be specific. "Create a webforJ project called InventorySystem with the sidemenu archetype and Spring Boot" beats "make an app".
• Ask for verification. Phrases like "verify against webforJ docs" or "check this CSS for bad --dwc-* tokens" nudge the assistant to use the tools instead of guessing.

If your assistant still answers without consulting the server, install the webforJ AI plugin - it ships matching Agent Skills that prompt the assistant to use the MCP tools automatically for webforJ tasks.

FAQ

Why isn't the AI assistant using the MCP server?

Why isn't the AI assistant using the MCP server?

Most assistants only reach for MCP when they think the question needs it. Two fixes:

1. Install the webforJ AI plugin, which pairs the server with Agent Skills that tell the assistant to use MCP for webforJ tasks.
2. Be explicit in your prompt: include "webforJ" in the question, and for stubborn cases say "use the webforJ MCP server to answer".

How to verify the MCP connection is working?

Use the MCP inspector:

npx @modelcontextprotocol/inspector

Then in the inspector, connect to https://mcp.webforj.com/mcp and explore the available tools.

How to report issues?

Open a ticket using the webforJ MCP issue template. Include the prompt, the expected result, and what you got.