Tool timeout within 10 seconds causing MCP server disconnect

[dino-su]

Description

The FastMCP server tools consistently timeout after 10 seconds of execution, triggering error code -32001 and forcing client disconnections:

2025-02-17T03:13:34.973Z [web-reader] [info] Message from client: {"jsonrpc":"2.0","method":"notifications/cancelled","params":{"requestId":27,"reason":"Error: MCP error -32001: Request timed out"}}

Impact:

1. All tool operations exceeding 10 seconds are forcefully cancelled
2. Server generates unhanded exceptions upon timeout
3. Client connections are terminated unexpectedly

Current Implementation

@mcp.tool(description="long running task") 
async def run(task: str) -> str:
    await asyncio.sleep(10)  # Task running for exactly 10 seconds
    return f"Task {task} completed."

Expected Behavior

• Configurable timeout thresholds for tool operations
• Graceful task cancellation with proper resource cleanup

Technical Questions

1. What is the FastMCP server's default timeout threshold?
2. How can the timeout threshold be configured per tool or globally?
3. What is the recommended approach for implementing long-running tools?

Environment

• FastMCP Server: v1.2.1

Reproduction Code

"""
Minimal reproduction case for FastMCP tool timeout issue.
"""

import asyncio
from mcp.server.fastmcp import FastMCP

mcp = FastMCP("timeout-demo")

@mcp.tool(description="long running task")
async def run(task: str) -> str:
    """
    Simulates a long-running task with 10-second duration.
    
    Args:
        task: Task identifier
    """
    await asyncio.sleep(10)
    return f"Task {task} completed."