Yep.
The two functions of interest are view.begin_edit(), which takes an optional command name and an optional dictionary of arguments, and view.end_edit(), which finishes the edit.
All actions done within an edit are grouped as a single undo action. Callbacks such as on_modified() and on_selection_modified() are called when the edit is finished.
It’s important to call view.end_edit() after each view.begin_edit(), otherwise the buffer will be in an inconsistent state. An attempt will be made to fix it automatically if the edit object gets collected, but that often doesn’t happen when you expect, and will result in a warning printed to the console. In other words, you should always bracket an edit in a try…finally block.
The command name passed to begin_edit() is used for repeat, macro recording, and for describing the action when undoing/redoing it. If you’re making an edit outside of a TextCommand, you should almost never supply a command name.
If you have created an edit object, and call a function that creates another one, that’s fine: the edit is only considered finished when the outermost call to end_edit() runs.
As well as grouping modifications, you can use edit objects for grouping changes to the selection, so they’re undone in a single step.