OpenAI MCP Server

Instructions

Implementation Reference

• src/mcp_server_openai/llm.py:10-33 (handler)

  Executes the core tool logic by calling the OpenAI chat completions API with predefined system prompts and the user query.

  async def ask_openai(self, query: str, model: str = "o3-mini") -> str:
      try:
          messages = [
              {
                  "role": "developer",
                  "content": "You are a helpful assistant that provides clear and accurate technical responses."
              },
              {
                  "role": "system",
                  "content": "Ensure responses are well-structured and technically precise."
              },
              {
                  "role": "user",
                  "content": query
              }
          ]
          response = await self.client.chat.completions.create(
              messages=messages,
              model=model
          )
          return response.choices[0].message.content
      except Exception as e:
          logger.error(f"Failed to query OpenAI: {str(e)}")
          raise

• src/mcp_server_openai/server.py:27-34 (schema)

  Input schema definition for the 'ask-openai' tool, specifying required 'query' and optional 'model'.

  inputSchema={
      "type": "object",
      "properties": {
          "query": {"type": "string", "description": "Ask assistant"},
          "model": {"type": "string", "default": "o3-mini", "enum": ["o3-mini", "gpt-4o-mini"]}
      },
      "required": ["query"]
  }

• src/mcp_server_openai/server.py:21-36 (registration)

  Registers the 'ask-openai' tool with MCP server including its name, description, and schema.

  @server.list_tools()
  async def handle_list_tools() -> list[types.Tool]:
      return [
          types.Tool(
              name="ask-openai",
              description="Ask my assistant models a direct question",
              inputSchema={
                  "type": "object",
                  "properties": {
                      "query": {"type": "string", "description": "Ask assistant"},
                      "model": {"type": "string", "default": "o3-mini", "enum": ["o3-mini", "gpt-4o-mini"]}
                  },
                  "required": ["query"]
              }
          )
      ]

• src/mcp_server_openai/server.py:38-55 (handler)

  MCP server tool call handler that dispatches 'ask-openai' calls to the LLMConnector and returns the response as TextContent.

  @server.call_tool()
  async def handle_tool_call(name: str, arguments: dict | None) -> list[types.TextContent]:
      try:
          if not arguments:
              raise ValueError("No arguments provided")
  
          if name == "ask-openai":
              response = await connector.ask_openai(
                  query=arguments["query"],
                  model=arguments.get("model", "o3-mini")
              )
              return [types.TextContent(type="text", text=response)]
  
          raise ValueError(f"Unknown tool: {name}")
      except Exception as e:
          logger.error(f"Tool call failed: {str(e)}")
          return [types.TextContent(type="text", text=f"Error: {str(e)}")]

• src/mcp_server_openai/llm.py:6-9 (helper)

  Helper class that initializes the AsyncOpenAI client used by the tool handler.

  class LLMConnector:
      def __init__(self, openai_api_key: str):
          self.client = AsyncOpenAI(api_key=openai_api_key)