I did it the other way round.
I have MTM/TFS refs in my tests. And then built myself a set of script extensions which use the MTM/TFS API to update the results of tests accordingly.
(The MTM/TFS API is pretty horrible. And still buggy. Updating runs - which is how it should be done - does not fully work. So I have that part stubbed at the moment and just update and individual case within a suite on a one-by-one basis. Not how it should be used, but until MS fix the API, it's my only option really.)
Auth tokens, URLs, project/suite/test ID are all specified as parameters at the start of, and during, the test run. As some things - like the project and auth token, only need updated once at the start of the run. Suites and cases can be constantly updated as TC runs through the suite.
Seems to work well enough so far. And can be switched off completely if I want to run outside MTM/TFS or on something we don't have in MTM/TFS.