Royal MCP for AI IDEs and CLIs
Royal MCP works with any MCP-compliant client, not just Claude Desktop. The same API-key-plus-URL pattern configures Cursor, VS Code, Windsurf, Claude Code, Zed, JetBrains AI Assistant, Continue, Cline, LM Studio, and any other IDE or CLI that speaks the Model Context Protocol.
How Royal MCP fits into any MCP client
IDE stands for Integrated Development Environment, the app a developer writes code in. VS Code, Cursor, Windsurf, JetBrains, and Zed are all AI-native IDEs. CLI is a command-line interface like Claude Code or Continue. MCP is Model Context Protocol, the open standard that lets any of these clients talk to Royal MCP running on your WordPress site.
MCP (Model Context Protocol) is an open standard. Any client that implements the protocol can talk to any MCP server, including Royal MCP running on your WordPress site. The client sends tool discovery and invocation requests to the MCP endpoint; Royal MCP responds with tool schemas and results.
Practically, connecting any AI IDE to Royal MCP comes down to three values: the server URL, your API key, and where to put the authorization header. The exact JSON shape varies by client, but the values are identical across all of them.
Supported IDEs and clients
claude mcp add royal-mcp .....vscode/mcp.json for server config.~/.continue/config.json.cline_mcp_settings.json.Authorization: Bearer <key>.Common configuration pattern
Most clients want the same three values: the server URL, your API key, and a header name. The JSON shape looks like this:
{
"name": "Royal MCP",
"url": "https://yoursite.com/wp-json/royal-mcp/v1/mcp",
"headers": {
"Authorization": "Bearer your-api-key-here"
}
}
For OAuth-capable clients, drop the headers block and the client will negotiate the access token via the OAuth metadata endpoint instead.
Every MCP client stores its config in a different place. Search the client’s docs for “MCP server” or “custom connector” to locate the right path. The URL and API key stay the same regardless of where you put them.
LM Studio setup
LM Studio added native MCP support in version 0.3.x. It expects a config file called mcp.json in your LM Studio config directory.
Config file location by OS:
- macOS:
~/.lmstudio/mcp.json - Windows:
%USERPROFILE%\.lmstudio\mcp.json - Linux:
~/.lmstudio/mcp.json
Option 1: direct HTTP (recommended for LM Studio 0.3.x+)
{
"mcpServers": {
"royal-mcp": {
"url": "https://example.com/wp-json/royal-mcp/v1/mcp",
"headers": {
"Authorization": "Bearer your-api-key-here"
}
}
}
}
Option 2: mcp-remote bridge (works on every LM Studio version)
The bridge form also works as a fallback for any stdio-only MCP client that can’t speak HTTP directly:
{
"mcpServers": {
"royal-mcp": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://example.com/wp-json/royal-mcp/v1/mcp",
"--header",
"Authorization:Bearer your-api-key-here"
]
}
}
}
After saving the file, restart LM Studio or press the “Reload MCP servers” button in newer versions. Royal MCP’s tools will appear in the model’s available-tools list.
- The
mcp-remoteform’s--headerargument has no space after the first colon (Authorization:Bearer VALUE, notAuthorization: Bearer VALUE). It’s a single CLI argument thatmcp-remotesplits on the first colon, not an HTTP header literal. The resulting HTTP request still goes out with the standardAuthorization: Bearer VALUEspacing. - The
mcp-remotebridge requires Node.js on the client machine fornpxto work. - The URL is the REST endpoint at
/wp-json/royal-mcp/v1/mcp, not the wp-admin page.
Ollama setup
Ollama itself does not have native MCP client support. It’s an LLM runtime, not an MCP host. To use Royal MCP “with Ollama,” you need a frontend that wraps Ollama and speaks MCP.
Recommended frontends that pair Ollama with MCP servers:
| Frontend | MCP support | Notes |
|---|---|---|
| Continue (VS Code / JetBrains) | Native | Same mcpServers JSON format. Set Ollama as the model provider in ~/.continue/config.json alongside the MCP server config. |
| OpenWebUI | Via mcpo plugin |
The mcpo bridge converts MCP to OpenAPI for OpenWebUI’s tool-calling layer. |
| LibreChat | Native (0.7.x+) | Configure MCP servers in librechat.yaml and select an Ollama-served model in the chat UI. |
| Cline | Native | Same JSON format as the other clients. Cline supports any OpenAI-compatible endpoint, including Ollama. |
The simplest “Ollama + Royal MCP” path: install Continue in VS Code or JetBrains, point it at your local Ollama install for the model, then use the same mcpServers config from the LM Studio section above.