Regarding Grahams response, the best way to handle this would be a recording application such as Cisco MediaSense or a 3rd party application which creates a SIP connecton and forks the audio for 100% call recording. You would configure this in CUCM in recording profile after creating the SIP profile.
As Graham said, he is correct that MedisSense is not the most robust QA solutions but it will bridge the gap. NICE (supported by Cisco) is a very expensive solution but it works very well then there are other low cost software solutions that handle SPAN based recording but you're limited to one local voice LAN but is not tied to the Cisco telephony solution except for JTAPI call control. A cheaper option is TelRex (cheaper and reliable), CallCopy, TASKE, etc but opt not to choose WFM add-ons.
Regarding Calabrio, it has nothing to do with the licensing model. Calabrio is a good recording solution and is deployed in most call centers I work with daily but Cisco Agent Desktop uses SPAN based recordings from the phone itself for monitoring and forks the audio path to the Cisco Supervisor Desktop making the request which minimizes inolvemnet with UCCX once the audio path is established. A conference call Calabrio would have to be able to interpret call control and support more than four audio paths (two calls at two audio paths each) to the supervisor desktop and likewise recording to UCCX. In addition, conference calling requires transcoding of which Calabrio won't support.
