Could you please clarify exactly what you want? Here are two possible interpretations I can offer an answer to:
An unstructured point cloud containing the matched 3D points from multiple cameras
This is currently not possible. You can use SaveImage and save one PCD for each camera’s point map and merge them afterwards. Or you can use SaveModel with the PointCloud parameter, but that currently only works for PLY files, although we are in the process of extending it to PCD (issue #6285).
A structured point map containing the matched 3D points from multiple cameras
This can be achieved by saving the result of the RenderPointMap command.