Effective system-level performance analysis in complex computing environments presents a hard problem. This is a problem shared by all platforms and solutions begin with pervasive instrumentation and measurement. This talk will describe the Performance Co-Pilot project and covers how python tools can be used to customise it for your own environment. A practical introduction will be given showing how to write python tools to export real-time performance metrics from arbitrary system components so that the data is available for live and retrospective analysis. A customised distributed analysis tool will then be built using that data.