To give a full answer we need to know (a) in what attribute these codes are stored and (b) if the attribute is a of type String, List or Set.

So, I'll assume the values are stored in a String-type attribute $MyString. As it is a String, we know there is only one value of $MyString per note. I'm testing in TB v5.12.2

Thus our agent's query is:

$MyString.contains("^K")

This matches any note where $MyString starts with a K, i.e. it is not a case sensitive match. so, on the agent's Rename dialog, tick the "Case sensitive" tick box to make the match case sensitive. Now, it won't match an empty attribute, "kden", "kDEN", or "BTHL". But it would match "KDE" or "KDEEN". To deal with the latter we should be able to use:

$MyString.contains("^K([A-Z]{3})$")

The parentheses () are to set a back-reference to part of the matched string so we can use it in the agent action:

$MyStringA = $1

The $1 value is it the 3 letters after the K captured in the () in the agent query (see more). I've used a new attribute $MyStringA so you can test the outcome without damaging the source data. Once happy, you can delete the $MyStringA attribute change the agent action to:

$MyString = $1

Here's a screengrab of my test:

It's probably worth mentioning the $ in the agent query.  In a regex ^ is start of string and $ is end of string. Earlier, I tested this:

$MyString.contains("^K([A-Z]{3})")

Which says "match from the start of a string, a capital K followed by exactly 3 uppercase letters". To my surprise, in my Tb test it matched the string KDEEN. What? That is K plus four uppercase letters. I think the issue is to do with greedy and non-greedy regex. That's a bit beyond me. The penny dropped, though, that I needed to address the rest of the string (if present) after the desired match. The simplest solution as your test case was for a fixed syntax of 4 character codes, was to add the $ regex character:

$MyString.contains("^K([A-Z]{3})$")

This now says "exactly match the whole string from start to finish, where it consists of a capital K followed by exactly 3 uppercase letters". In the test this now excludes KKK (to few characters) and KDEEN (too many).

I thought it worth adding the extra in as it's an issue I often trip over as an occasional regex user in TB where I only want to match the beginning of a string and may efforts fail because I don't take into account matching - or not of the remaining part of the string.

Thanks for the correction. The fact one is doing a 'contains' should tip one off that you need to use ^ and $, with appropriate regex in between, to ensure a full string match.  

The is no method to do an '=' match, something like:

$SomeStringAttr = regex("K[A-Z]{3}")

and have it match only 4 character values that match the regex. To be clear, there is no such TB operator as regex(), it's just a made-up example.