MCP Server

gllm-inference | Tutorial: MCP Server | API Reference

Supported by: OpenAILMInvoker

What is MCP Server?

MCP server is a native tool that allows the language model to use remote MCP servers to give models new capabilities. When it's enabled, MCP calls are stored in the outputs attribute of the LMOutput object and can be accessed via the mcp_calls property.

MCP server tool can be enabled with several options:

import asyncio
from gllm_inference.lm_invoker import OpenAILMInvoker
from gllm_inference.model import OpenAILM
from gllm_inference.schema import NativeTool, NativeToolType

SERVER_URL = "https://mcp.deepwiki.com/mcp"
SERVER_NAME = "deepwiki"

# Option 1: as dictionary 
mcp_server_tool = {"type": "mcp_server", "url": SERVER_URL, "name": SERVER_NAME, **kwargs}
# Option 2: as native tool object 
mcp_server_tool = NativeTool.mcp_server(url=SERVER_URL, name=SERVER_NAME, kwargs)

lm_invoker = OpenAILMInvoker(OpenAILM.GPT_5_NANO, tools=[mcp_server_tool])

Let's try it to utilize the MCP server through our LM invoker!

query = """
List all transport protocols supported by the 2025-03-26 version of 
the MCP spec (modelcontextprotocol/modelcontextprotocol) in bullet points"
"""
output = asyncio.run(lm_invoker.invoke(query))
for item in output.outputs:
    print(f"=== Output item: {item.type!r} ===\n{item.output}\n")

Output:

=== Output item: 'mcp_call' ===
id='mcp_0bb324738ba65b790069649beeafe08193a485b62fee44f228' 
server_name='deepwiki' 
tool_name='read_wiki_structure' 
args={'repoName': 'modelcontextprotocol/modelcontextprotocol'} 
output='Available pages for modelcontextprotocol/modelcontextprotocol...'

=== Output item: 'mcp_call' ===
id='mcp_0bb324738ba65b790069649bf2395481939dac9e53eb98818d' 
server_name='deepwiki' 
tool_name='read_wiki_contents' 
args={'repoName': 'modelcontextprotocol/modelcontextprotocol'} 
output='# Page: Overview\n\n# Overview\n\n<details>\n<summary>...'

=== Output item: 'mcp_call' ===
id='mcp_0bb324738ba65b790069649bf81cf08193b7066a98abaf56e1' 
server_name='deepwiki' 
tool_name='ask_question' 
args={
    'repoName': 'modelcontextprotocol/modelcontextprotocol', 
    'question': 'List all transport protocols supported by the 2025-03-26 MCP spec (modelcontextprotocol/modelcontextprotocol). (From the Transport Layer section 2.3.)',
} 
output='The 2025-03-26 Model Context Protocol (MCP) specification supports...'

=== Output item: 'text' ===
- STDIO Transport
- Streamable HTTP Transport (HTTP POST with optional Server-Sent Events for streaming)