Read the documentation!

mcp-server-llmling is a server for the Machine Chat Protocol (MCP) that provides a YAML-based configuration system for LLM applications.

LLMLing, the backend, provides a YAML-based configuration system for LLM applications. It allows to set up custom MCP servers serving content defined in YAML files.

• Static Declaration: Define your LLM's environment in YAML - no code required
• MCP Protocol: Built on the Machine Chat Protocol (MCP) for standardized LLM interaction
• Component Types:
  • Resources: Content providers (files, text, CLI output, etc.)
  • Prompts: Message templates with arguments
  • Tools: Python functions callable by the LLM

The YAML configuration creates a complete environment that provides the LLM with:

• Access to content via resources
• Structured prompts for consistent interaction
• Tools for extending capabilities

• Load and manage different types of resources:
  • Text files (PathResource)
  • Raw text content (TextResource)
  • CLI command output (CLIResource)
  • Python source code (SourceResource)
  • Python callable results (CallableResource)
  • Images (ImageResource)
• Support for resource watching/hot-reload
• Resource processing pipelines
• URI-based resource access

• Register and execute Python functions as LLM tools
• Support for OpenAPI-based tools
• Entry point-based tool discovery
• Tool validation and parameter checking
• Structured tool responses

• Static prompts with template support
• Dynamic prompts from Python functions
• File-based prompts
• Prompt argument validation
• Completion suggestions for prompt arguments

• Stdio-based communication (default)
• Server-Sent Events (SSE) for web clients
• Support for custom transport implementations

Add LLMLing as a context server in your settings.json:

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

Configure LLMLing in your claude_desktop_config.json:

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

Start the server directly from command line:

# Latest version
uvx mcp-server-llmling@latest

from llmling import RuntimeConfig
from mcp_server_llmling import LLMLingServer

async def main() -> None:
    async with RuntimeConfig.open(config) as runtime:
        server = LLMLingServer(runtime, enable_injection=True)
        await server.start()

asyncio.run(main())

from llmling import RuntimeConfig
from mcp_server_llmling import LLMLingServer

async def main() -> None:
    async with RuntimeConfig.open(config) as runtime:
        server = LLMLingServer(
            config,
            transport="sse",
            transport_options={
                "host": "localhost",
                "port": 8000,
                "cors_origins": ["http://localhost:3000"]
            }
        )
        await server.start()

asyncio.run(main())

resources:
  python_code:
    type: path
    path: "./src/**/*.py"
    watch:
      enabled: true
      patterns:
        - "*.py"
        - "!**/__pycache__/**"

  api_docs:
    type: text
    content: |
      API Documentation
      ================
      ...

tools:
  analyze_code:
    import_path: "mymodule.tools.analyze_code"
    description: "Analyze Python code structure"

toolsets:
  api:
    type: openapi
    spec: "https://api.example.com/openapi.json"
    namespace: "api"

The server is configured through a YAML file with the following sections:

global_settings:
  timeout: 30
  max_retries: 3
  log_level: "INFO"
  requirements: []
  pip_index_url: null
  extra_paths: []

resources:
  # Resource definitions...

tools:
  # Tool definitions...

toolsets:
  # Toolset definitions...

prompts:
  # Prompt definitions...

The server implements the MCP protocol which supports:

1. Resource Operations

   • List available resources
   • Read resource content
   • Watch for resource changes

2. Tool Operations

   • List available tools
   • Execute tools with parameters
   • Get tool schemas

3. Prompt Operations

   • List available prompts
   • Get formatted prompts
   • Get completions for prompt arguments

4. Notifications

   • Resource changes
   • Tool/prompt list updates
   • Progress updates
   • Log messages