As far as I know Mock doesn’t provide a way to achieve what you want via assert_called_with. You could access the call_args and call_args_list members and perform the assertions manually.

However the is a simple (and dirty) way of achieving almost what you want. You have to implement a class whose __eq__ method always returns True:

def Any(cls):
    class Any(cls):
        def __eq__(self, other):
            return True
    return Any()

Using it as:

In [14]: caller = mock.Mock(return_value=None)


In [15]: caller(1,2,3, arg=True)

In [16]: caller.assert_called_with(Any(int), Any(int), Any(int), arg=True)

In [17]: caller.assert_called_with(Any(int), Any(int), Any(int), arg=False)
---------------------------------------------------------------------------
AssertionError                            Traceback (most recent call last)
<ipython-input-17-c604faa06bd0> in <module>()
----> 1 caller.assert_called_with(Any(int), Any(int), Any(int), arg=False)

/usr/lib/python3.3/unittest/mock.py in assert_called_with(_mock_self, *args, **kwargs)
    724         if self.call_args != (args, kwargs):
    725             msg = self._format_mock_failure_message(args, kwargs)
--> 726             raise AssertionError(msg)
    727 
    728 

AssertionError: Expected call: mock(0, 0, 0, arg=False)
Actual call: mock(1, 2, 3, arg=True)

As you can see it only checks for the arg. You have to create subclasses of int, otherwise the comparisons wont work¹. However you still have to provide all the arguments. If you have many arguments you might shorten your code using tuple-unpacking:

In [18]: caller(1,2,3, arg=True)

In [19]: caller.assert_called_with(*[Any(int)]*3, arg=True)

Except for this I cannot think of a way to avoid passing all parameters to assert_called_with and work it as you intend.

---

The above solution can be extended to check for types of other arguments. For example:

In [21]: def Any(cls):
    ...:     class Any(cls):
    ...:         def __eq__(self, other):
    ...:             return isinstance(other, cls)
    ...:     return Any()

In [22]: caller(1, 2.0, "string", {1:1}, [1,2,3])

In [23]: caller.assert_called_with(Any(int), Any(float), Any(str), Any(dict), Any(list))

In [24]: caller(1, 2.0, "string", {1:1}, [1,2,3])

In [25]: caller.assert_called_with(Any(int), Any(float), Any(str), Any(dict), Any(tuple))
---------------------------------------------------------------------------
AssertionError                            Traceback (most recent call last)
<ipython-input-25-f607a20dd665> in <module>()
----> 1 caller.assert_called_with(Any(int), Any(float), Any(str), Any(dict), Any(tuple))

/usr/lib/python3.3/unittest/mock.py in assert_called_with(_mock_self, *args, **kwargs)
    724         if self.call_args != (args, kwargs):
    725             msg = self._format_mock_failure_message(args, kwargs)
--> 726             raise AssertionError(msg)
    727 
    728 

AssertionError: Expected call: mock(0, 0.0, '', {}, ())
Actual call: mock(1, 2.0, 'string', {1: 1}, [1, 2, 3])

however this doesn’t allow arguments that can be, for example, both an int or a str. Allowing multiple arguments to Any and using multiple-inheritance wont help. We can solve this using abc.ABCMeta

def Any(*cls):
    class Any(metaclass=abc.ABCMeta):
        def __eq__(self, other):
            return isinstance(other, cls)
    for c in cls:
        Any.register(c)
    return Any()

Example:

In [41]: caller(1, "ciao")

In [42]: caller.assert_called_with(Any(int, str), Any(int, str))

In [43]: caller("Hello, World!", 2)

In [44]: caller.assert_called_with(Any(int, str), Any(int, str))

---

¹ I used the name Any for the function since it is “used as a class” in the code. Also any is a built-in…