MCP Ollama Server

Instructions

Args:
    model: Name of the model to use (e.g., 'llama2')
    question: The question to ask the model

Implementation Reference

• src/mcp_ollama/server.py:73-91 (handler)

  The main handler function for the 'ask_model' tool. It is decorated with @mcp.tool() which registers it in the MCP server. The function takes a model name and question, sends a chat request to Ollama, and returns the response content. Includes input schema via type hints and docstring.

  @mcp.tool()
  async def ask_model(model: str, question: str) -> str:
      """Ask a question to a specific Ollama model
  
      Args:
          model: Name of the model to use (e.g., 'llama2')
          question: The question to ask the model
      """
      try:
          response = ollama.chat(
              model=model,
              messages=[{
                  'role': 'user',
                  'content': question
              }]
          )
          return response['message']['content']
      except Exception as e:
          return f"Error querying model: {str(e)}"

• src/mcp_ollama/server.py:73-73 (registration)

  The @mcp.tool() decorator registers the ask_model function as an MCP tool.

  @mcp.tool()

• src/mcp_ollama/server.py:74-80 (schema)

  The function signature and docstring define the input schema (model: str, question: str) and output (str).

  async def ask_model(model: str, question: str) -> str:
      """Ask a question to a specific Ollama model
  
      Args:
          model: Name of the model to use (e.g., 'llama2')
          question: The question to ask the model
      """