This proposal look preaty close to pep-463: http://legacy.python.org/dev/peps/pep-0463/, but in assertion context.

Now python test libraries have different aproach for assertions, some try use own implementations, for example, for equality `assertEqual` (`assertEqulas`), `eq_`, some use python `assert` keyword.

`assert` work fine for variables, but requred verbose wrappers for exception raising, for example:

>>> try:
...   do_raise_code()
... except Exception:
...   assert False

Test libraries already have self implementations but it still not pure python like `assert`:

unittest: `self.assertRaises`
py.test: `pytest.raises`
nose: `nose.tools.raises` or `nose.tools.assert_raises`

I propose add pure python implementation for this case because it enough popular, for example:

>>> assert do_raise_code() raises  # ok if `do_raise_code` raise any exception
>>> assert do_raise_code() raises, 'text message'  # ok if `do_raise_code` raise any exception with message
>>> assert do_raise_code() raises Exception  # ok if `do_raise_code` raise specific exception
>>> assert do_raise_code() raises Exception, 'text message'  # ok if `do_raise_code` raise specific exception with message
>>> assert do_raise_code() raises (TypeError, ValueError)  # ok if `do_raise_code` raise one of specific exceptions
>>> assert do_raise_code() raises (TypeError, ValueError), 'text message'  # ok if `do_raise_code` raise one of specific exceptions with message

Test libraries can use tham raises implementations as decorator, this proposal currently ignore similar behaviour.
Test libraries can use them raises implementations as context, this propasal currently ignore similar behaviour.
`unittest` module also has `assertRaisesRegex` method, this proposal currently ignore similar behaviour.
`unittest` module also has `assertWarns` and `assertLogs`, this proposal currently ignore similar behaviour.
Also this proposal currently ignore any access to exception object and it fields.

The best way to get language changes made is to first propose them on the python-ideas mailing list (https://docs.python.org/devguide/faq.html#suggesting-changes).

Agreed, but there are a few problems with the proposals:
1) it requires the addition of a new keyword (raises), and that will break any program that uses it as an identifier;
2) the use case is arguably limited;
3) assertRaises is consistent with the other assert* methods, and it also provides way to inspect the exception, whereas IIUC what you are proposing does not (and also doesn't replace assertRaisesRegex);
4) I don't find your proposal simpler or more readable than what we have now:

In other words there are quite a few problems for little benefit, so even if you decide to bring this up to python-ideas, it's very unlikely that it will be accepted in its current form.

Thanks for reply, I really missed many cases. I will try look deeper.

See also https://mail.python.org/pipermail/python-ideas/2014-June/028206.html for a list of things to keep in mind while adding new features (especially keywords).