MCP server fails to start: FastMCP.tool() does not accept `timeout` kwarg

[JaysonTurkRP]

Bug Description

The MCP server fails to start because FastMCP.tool() does not support the timeout keyword argument. Several tool registrations pass timeout= to the @mcp.tool() decorator, which causes a TypeError at import time.

Error

TypeError: FastMCP.tool() got an unexpected keyword argument 'timeout'

Affected Files

• databricks_mcp_server/tools/pipelines.py line 156: @mcp.tool(timeout=300)
• databricks_mcp_server/tools/pdf.py line 14: @mcp.tool(timeout=300.0)
• databricks_mcp_server/tools/pdf.py line 104: @mcp.tool(timeout=60.0)

Environment

• mcp: 1.26.0
• fastmcp: 2.14.5
• OS: Windows 11
• Python: 3.12 (ai-dev-kit venv)
• AI Dev Kit version: 0.1.6

Root Cause

FastMCP.tool() in fastmcp 2.14.5 does not accept a timeout parameter. The decorator signature only supports name, description, annotations, and tags. The timeout kwarg was likely intended for a newer or different version of FastMCP that hasn't been released yet, or for a custom wrapper.

Workaround

Adding kwargs.pop("timeout", None) to the patched_tool() function in server.py (the existing Windows compatibility patch in _patch_tool_decorator_for_windows) strips the unsupported kwarg before it reaches FastMCP.tool():

@functools.wraps(original_tool)
def patched_tool(fn=None, *args, **kwargs):
    # Strip kwargs not supported by FastMCP.tool()
    kwargs.pop("timeout", None)
    # Handle @mcp.tool("name") — returns a decorator
    if fn is None or isinstance(fn, str):
        ...

Suggested Fix

Either:

1. Remove timeout= from the @mcp.tool() decorators in pipelines.py and pdf.py
2. Pin a fastmcp version that supports timeout in pyproject.toml / requirements.txt
3. Add the kwargs.pop("timeout", None) workaround to server.py's existing patched_tool() function

Reproducibility

This breaks on every fresh install or update of ai-dev-kit v0.1.6 when using fastmcp 2.14.5. The MCP server cannot start at all — no tools are available.