MCP - Model Context Protocol - Server Support

[julweber]

Hi guys,

first up thanks for your great work. I just found out about Newelle and i already love it after an hour of usage.

I have a feature recommendation that is a wish for many users, i guess.

MCP Server Configuration for Newelle AI

I propose adding support for configuring Model Context Protocol (MCP) servers through a JSON configuration file, enabling seamless integration with external services.

Recommended Configuration Format Example

It would make sense to follow the "standard" formats approaching in the context of Coding tools.

{
  "mcpServers": {
    "context7": {
      "type": "streamable-http",
      "url": "https://mcp.context7.com/mcp"
    }
  }
}

This format follows the established pattern from tools like Roo Code and provides:
See here for example: Context 7 Roo Code configuration or here: Roo Code MCP Configuration

• Type specification for different MCP server implementations (sse, streamable-http (http), stdio)
• Simple URL configuration for service endpoints
• Named server definitions for easy reference and management

Feature Benefits

This integration allows Newelle AI to:

• Connect any MCP-compatible server directly into chat contexts
• Access external data sources and services in real-time
• Extend LLM capabilities without custom development
• Support the growing MCP ecosystem standard

Implementation

• Users can define their MCP servers in a mcp.json configuration file.
• The application will automatically discover and integrate these services, making them available within conversation contexts. - The UI provides a configuration window to activate and deactivate selected MCP servers for the current chat.
• The general configuration UI allows displaying and managing and selecting a set of mcp servers that are active for new chats per default.

I encourage anybody to add onto this specification and make this as easy as possible to implement.

Cheers and thanks for the great work

[FrancescoCaracciolo]

It's the next big thing we are adding, see #223