#benwerdKnown does attempt to be LTI compliant, and there will be more automated tests added around that; the idea is also to be relatively agnostic as to which data clients want to consume. JSON-LD isn't my favorite, but it's used by a few downstream tools, and doesn't add much overhead.