When generating tools, function parameters should allow annotations(Annotated args) to propagate to the inputSchema of each tool.

[DmitriyLeybel]

Is your feature request related to a problem? Please describe.
Currently, the parameters are typically best described within the docstring. This is a subpar approach and can be further refined to create a better inputSchema when the tool list is emitted.

e.g.

@mcp.tool()
def generate_random_number(min_value: int = 1, max_value: int = 100) -> int:
    """
    Generate a random number between min_value and max_value (inclusive).
    
    Args:
        min_value: Minimum value (default: 1)
        max_value: Maximum value (default: 100)
        
    Returns:
        A random integer
    """
    return random.randint(min_value, max_value)

Describe the solution you'd like
Instead, we can use the Annotated feature of Python and pass in the appropriate metadata into the function signature.

@mcp.tool()
def generate_random_number(
    min_value: Annotated[int, {"description": "The inclusive minimum value for the random number",
                                               "minimum": 0}] = 1,
    max_value: Annotated[int, {"description": "The inclusive maximum value for the random number",
                                               "maximum": 1000000000}] = 100,
) -> int:
    """
    Generate a random number between min_value and max_value (inclusive).
    """
    if min_value > max_value:
        raise ValueError("min_value must be less than or equal to max_value")
    return random.randint(min_value, max_value)

Describe alternatives you've considered
Considered Pydantic fields, but that's unnecessary overhead for simple cases.

---

Seems like a fairly simple addition to the func_metadata method.
https://github.com/modelcontextprotocol/python-sdk/blob/main/src/mcp/server/fastmcp/utilities/func_metadata.py#L105