In this paper, we show that the LZ77 factorization of a text T ∈ Σn can be computed in O(R log n) bits of working space and O(n log R) time, R being the number of runs in the Burrows-Wheeler transform of T (reversed). For (extremely) repetitive inputs, the working space can be as low as O(log n) bits: exponentially smaller than the text itself. Hence, our result finds important applications in the construction of repetition-aware self-indexes and in the compression of repetitive text collections within small working space.