Tyler Forge™ MCP Server

Get instant access to Tyler Forge™ web component documentation directly in AI clients. Ask questions, generate code examples, and explore design system details.

What You Get

Tyler Forge™ is Tyler Technologies' comprehensive web component library and design system. This tool brings that documentation into your AI tools so you can:

• Discover components - Find the right component for your needs
• Generate code - Get framework-specific examples (Angular, React, Vue, Svelte)
• Validate APIs - Verify that component properties, attributes, events, and other APIs are correctly used
• Access design tokens - Use colors, spacing, typography consistently
• Learn best practices - Get implementation guidance and accessibility tips

How does it work?

This MCP server provides a set of declarative resources and tools for accessing Tyler Forge documentation. It uses a progressive disclosure model to minimize complexity while maximizing accuracy.

Context-Aware

When you start the MCP server in your AI client, it automatically detects the currently installed version of Tyler Forge (if any) in your project. This allows it to provide version-specific documentation. If no version is detected, it defaults to the latest stable release that is bundled within this MCP server.

Progressive Disclosure

The MCP server is designed to minimize cognitive load by exposing only the most essential resources and tools upfront. More detailed information can be accessed through specific tools or follow-up queries. This keeps control of the token usage while still providing comprehensive access to the documentation when needed.

Documentation Source

All documentation is either pulled from the Custom Elements Manifest (CEM) files included with each Tyler Forge package, as well as information that was sourced from the documentation site at https://forge.tylerdev.io.

Setup

Claude Code

bashCopy code
claude mcp add -t stdio -s [scope] forge -- npx -y @tylertech/forge-mcp@latest

The [scope] must be user, project or local. See the Claude Code MCP documentation for more information.

Codex CLI

Add the following to your config.toml (which defaults to ~/.codex/config.toml, but refer to the configuration documentation for more advanced setups):

tomlCopy code
[mcp_servers.forge]
command = "npx"
args = ["-y", "@tylertech/forge-mcp@latest"]

Gemini CLI

bashCopy code
gemini mcp add -t stdio -s [scope] forge npx -y @tylertech/forge-mcp@latest

The [scope] must be user, project or local.

VS Code

• Open the Command Palette (Cmd+Shift+P or Ctrl+Shift+P)
• Choose MCP: Add Server...
• Choose Command (stdio)
• Enter npx -y @tylertech/forge-mcp@latest
• Name the server forge
• Choose if you want to use it as a Global or Workspace MCP server

See the VS Code MCP docs for more info.

Manual Configuration

Add the Forge MCP server to the .vscode/mcp.json configuration file in your project, or use the Command Palette to add it via "MCP: Add Server...".

jsonCopy code
{
  "servers": {
    "forge": {
      "command": "npx",
      "args": ["-y", "@tylertech/forge-mcp@latest"],
      "type": "stdio"
    }
  }
}

Note: You may need to start the server manually after adding it to the config file. Use the MCP: List Servers command from the Command Palette.

Claude Desktop

• Open Settings
• Choose Developer
• Click on Edit Config

This will open your file explorer to the directory where the claude_desktop_config.json file lives. Edit the file to include the following configuration:

jsonCopy code
{
  "mcpServers": {
    "forge": {
      "command": "npx",
      "args": ["-y", "@tylertech/forge-mcp@latest"]
    }
  }
}

Best Practices

When using any AI-powered tool, it's important to follow best practices to ensure effective usage and accurate results:

• Be Specific: Clearly define your questions or requests to get the most relevant information. Using vague prompts can lead to inaccurate or incorrect responses.
• Plan Ahead: Spend time outlining and planning your prompts, and break things down into smaller tasks. This can significantly improve the quality of the responses.
• Validate: Always cross-check critical information with official documentation or trusted sources. AI output may not always be accurate.
• Provide Context: When writing prompts, provide relevant context to help the AI understand your needs better.
• Use Examples: When possible, provide examples or screenshots of what you're looking for to guide the AI's responses.
• Limit Scope: Break down complex queries into smaller, manageable parts for better clarity and responses. For lengthy conversations, periodically compact or clear context to maintain relevance.

If there's one thing to take away, it's to be specific and spend time up front planning and structuring your prompts. This will lead to more accurate and useful results. In the end, these tools are here to assist you, and they can get things wrong even with the best prompts and context. Always verify and review the output for accuracy.

Important Note

ALWAYS validate the output from any AI tool for accuracy. While this MCP server is designed to provide accurate information, it may not always be perfect and LLMs are known to hallucinate at times. Always cross-check critical information with the official Tyler Forge™ documentation or trusted sources.

Capabilities

Prompts

forge_mode

Sets some baseline rules when running Forge-specific tasks/queries.

You can trigger this prompt in your AI client by specifying the prompt name (if supported), typically via a slash command or prompt selection. The forge_mode prompt accepts a task parameter where you can describe your specific task/query. This helps guide the LLM to use the correct MCP server tools, as well as includes some rules and best practices when outputting Tyler Forge-related code.

Tools

The following tools are available for interacting with Tyler Forge documentation. Each tool uses declarative parameters to provide comprehensive functionality while minimizing the cognitive load for AI clients:

Resources

The following streamlined resources are available for reference via a progressive disclosure model:

Note: Granular component resources (/properties, /methods, /events, etc.) have been removed to reduce complexity. Use the tools for granular access to specific component information.

Development

Commands:

bashCopy code
pnpm run dev    # Watch mode
pnpm run debug  # Test with MCP inspector
pnpm run build  # Build for production

Contributing

Contributions are welcome! Please see the CONTRIBUTING guide for details.

If you find any issues or have suggestions for improvements, please open an issue or submit a pull request.

License

Apache-2.0 License - see LICENSE file for details.

Part of the Tyler Technologies Open Source initiative