Pydantic integration not working.

[Javedgouri]

Description

As I have created tool request body with Pydantic model .Unable to view Pydantic model request body in MCP Inspector (stdio)

I have attached code below and screenshot of mcp inscpector

Example Code

from fastmcp import FastMCP

from models import (
    CalculateRequest, KmToMilesRequest,
    CurrencyConversionRequest, StockPriceRequest
)

# Initialize the MCP server with a name
mcp = FastMCP("Tools Server")


@mcp.tool()
def calculate(request: CalculateRequest) -> int | float:
    """Perform basic arithmetic operations."""
    operations = {
        "add": lambda x, y: x + y,
        "subtract": lambda x, y: x - y,
        "multiply": lambda x, y: x * y,
        "divide": lambda x, y: "Error: Cannot divide by zero" if y == 0 else x / y
    }
    result = operations.get(request.operation, lambda x, y: None)(request.number1, request.number2)
    return result


@mcp.tool()
def convert_km_to_miles(request: KmToMilesRequest) -> float:
    """Convert kilometers to miles."""
    miles = round(request.km * 0.621371, 4)
    return float(miles)


@mcp.tool()
def convert_currency(request: CurrencyConversionRequest) -> float:
    """Convert currency."""
    dummy_rates = {
        ("USD", "INR"): 83.0,
        ("INR", "USD"): 0.012,
        ("EUR", "USD"): 1.1,
    }
    rate = dummy_rates.get((request.from_currency.upper(), request.to_currency.upper()), 1.0)
    converted_amount = round(request.amount * rate, 2)

    return float(converted_amount)


@mcp.tool()
def get_stock_price(request: StockPriceRequest) -> str:
    """Get stock price."""
    dummy_prices = {
        "AAPL": 173.54,
        "GOOGL": 134.21,
        "TSLA": 192.35,
    }
    price = dummy_prices.get(request.symbol.upper(), 100.0)
    message = f"{request.symbol.upper()} is trading at ${price}"
    return message


####### models.py 
from typing import Literal, Optional, Union
from pydantic import BaseModel, Field


# Request Models
class CalculateRequest(BaseModel):
    """Request model for performing basic arithmetic operations."""
    operation: Literal['add', 'subtract', 'multiply', 'divide'] = Field(
        description="The arithmetic operation to perform"
    )
    number1: int = Field(description="First operand for the calculation")
    number2: int = Field(description="Second operand for the calculation")


class KmToMilesRequest(BaseModel):
    """Request model for converting kilometers to miles."""
    km: float = Field(description="Distance in kilometers to be converted")


class CurrencyConversionRequest(BaseModel):
    """Request model for currency conversion."""
    amount: float = Field(description="Amount of currency to convert")
    from_currency: str = Field(description="Source currency code (e.g., USD, EUR)")
    to_currency: str = Field(description="Target currency code (e.g., INR, USD)")

class StockPriceRequest(BaseModel):
    """Request model for retrieving stock price."""
    symbol: str = Field(description="Stock symbol (e.g., AAPL, GOOGL, TSLA)")

Version Information

factmcp version - 2.2.2

Additional Context

No response

[strawgate]

I believe this is an mcp inspector issue, if you look at the schema the schema is correct and the LLM is able to interact with it

[jlowin]

I'm able to replicate but I'm inclined to agree this is a Inspector issue. I think it might be that the inspector doesn't properly resolve $ref references in JsonSchema (which are always produced when Pydantic objects are used as arguments). I am checking if it's possible to flatten them with jsonref but this might lead to performance regressions elsewhere.

[Javedgouri]

@jlowin This is input schema of one of the tool which doesn't seems correct { "$defs": { "CalculateRequest": { "description": "Request model for performing basic arithmetic operations.", "properties": { "number1": { "description": "First operand for the calculation", "title": "Number1", "type": "integer" }, "number2": { "description": "Second operand for the calculation", "title": "Number2", "type": "integer" }, "operation": { "description": "The arithmetic operation to perform", "enum": ["add", "subtract", "multiply", "divide"], "title": "Operation", "type": "string" } }, "required": ["operation", "number1", "number2"], "title": "CalculateRequest", "type": "object" } }, "properties": { "request": { "$ref": "#/$defs/CalculateRequest" } }, "required": ["request"], "title": "calculateArguments", "type": "object" }

I am getting this issue while calling this tool with client | fastmcp.exceptions.ClientError: Error executing tool calculate: 1 validation error for calculateArguments | request | Field required [type=missing, input_value={'operation': 'add', 'number1': 5, 'number2': 10}, input_type=dict]

[strawgate]

That schema looks correct to me, what part looks incorrect?

[jlowin]

The validation error is because you are supplying the wrong arguments, you defined the tool with a request key:

async with Client(mcp) as client:
    result = await client.call_tool(
        "calculate",
        {
            "request": {
                "operation": "add",
                "number1": 5,
                "number2": 10,
            }
        },
    )
    print(result)

Above will print [TextContent(type='text', text='15', annotations=None)]

[jlowin]

I believe this is either an intentional (see https://github.com/modelcontextprotocol/inspector/pull/284 and https://github.com/modelcontextprotocol/inspector/issues/331) or accidental (https://github.com/modelcontextprotocol/inspector/issues/332) decision in the MCP Inspector. I can confirm that FastMCP is producing the correct tool schemas with complex objects and agents like Pydantic AI agents call them correctly.