MCP Server Llmling

🚀 MCP-Server-Llmling Project Introduction

This project, MCP-Server-Llmling, provides a comprehensive set of features for resource management, tool execution, prompt handling, and communication. It offers a flexible and efficient way to interact with LLMs and manage related resources.

✨ Features

1. Resource Management

• Load and manage various types of resources:
  • Text files (PathResource)
  • Plain text content (TextResource)
  • CLI command outputs (CLIResource)
  • Python source code (SourceResource)
  • Python callable results (CallableResource)
  • Images (ImageResource)
• Support real-time resource monitoring and hot reloading.
• Resource processing pipelines.
• URI-based resource access.

2. Tool System

• Register and execute Python functions as LLM tools.
• Support OpenAPI-based tools.
• Discover base tools from entry points.
• Validate tools and check parameters.
• Provide structured tool responses.

3. Prompt Management

• Support static prompts with template support.
• Generate dynamic prompts from Python functions.
• Use file-based prompts.
• Validate prompt parameters.
• Provide completion suggestions for prompt parameters.

4. Multiple Transport Options

• Communicate based on standard input/output (default).
• Use Server-Sent Events (SSE) for web clients.
• Support custom transport implementations.

💻 Usage Examples

Using in Zed Editor

Configure LLMLing as the context server in settings.json:

{
  "context_servers": {
    "llmling": {
      "command": {
        "env": {},
        "label": "llmling",
        "path": "uvx",
        "args": [
          "mcp-server-llmling",
          "start",
          "path/to/your/config.yml"
        ]
      },
      "settings": {}
    }
  }
}

Using in Claude Desktop

Configure LLMLing in claude_desktop_config.json:

{
  "mcpServers": {
    "llmling": {
      "command": "uvx",
      "args": [
        "mcp-server-llmling",
        "start",
        "path/to/your/config.yml"
      ],
      "env": {}
    }
  }
}

Manually Starting the Server

Start the server directly from the command line:

# Latest version
uvx mcp-server-llmling@latest

1. Programming Usage

from llmling import RuntimeConfig
from mcp_server_llmling import LLMLingServer

async def main() -> None:
    async with await RuntimeConfig.get_instance():
        server = await LLMLingServer.create()
        await server.start()

if __name__ == "__main__":
    asyncio.run(main())

2. Resource Management Example

from llmling.resources import ResourceManager

async def main() -> None:
    async with await RuntimeConfig.get_instance():
        resource_manager = ResourceManager()
        # Load resource
        await resource_manager.load("path/to/resource")
        # Get resource
        resource = await resource_manager.get("resource_id")

if __name__ == "__main__":
    asyncio.run(main())

3. Tool Management Example

from llmling.tools import ToolManager

async def main() -> None:
    async with await RuntimeConfig.get_instance():
        tool_manager = ToolManager()
        # Register tool
        await tool_manager.register("tool_id", lambda: "result")
        # Get tool
        tool = await tool_manager.get("tool_id")

if __name__ == "__main__":
    asyncio.run(main())

4. Prompt Management Example

from llmling.prompts import PromptManager

async def main() -> None:
    async with await RuntimeConfig.get_instance():
        prompt_manager = PromptManager()
        # Load prompt
        await prompt_manager.load("path/to/prompt_file")
        # Get prompt
        prompt = await prompt_manager.get("prompt_id")

if __name__ == "__main__":
    asyncio.run(main())

5. SSE Client Example

import asyncio
import json
import websockets

async def client():
    async with websockets.connect('ws://localhost:8000') as websocket:
        await websocket.send(json.dumps({
            "type": "subscribe",
            "id": "resource_update"
        }))
        try:
            while True:
                message = await websocket.recv()
                print(f"Received: {message}")
        except websockets.exceptions.ConnectionClosedError:
            print("Connection closed")

if __name__ == "__main__":
    asyncio.get_event_loop().run_until_complete(client())

📚 Documentation

Configuration Examples

Resource Configuration

resources:
  - id: "resource_id"
    type: "PathResource"
    path: "path/to/resource"
  - id: "another_resource"
    type: "TextResource"
    text: "sample text content"

Tool Configuration

tools:
  - id: "tool_id"
    type: "FunctionTool"
    function: "module.function"
  - id: "another_tool"
    type: "OpenAIChatCompletion"
    api_key: "your_api_key"

Prompt Configuration

prompts:
  - id: "prompt_id"
    type: "FilePrompt"
    path: "path/to/prompt_file"
  - id: "another_prompt"
    type: "StringPrompt"
    content: "sample prompt text"

Supported Tools

• Zed Editor: Configure in settings.json.
• Claude Desktop: Configure in claude_desktop_config.json.
• Other Tools: Integrate through custom configuration.