We show that direct feedback based on quantum jump detection can be used to generate entangled steady states. We present a strategy that is insensitive to detection inefficiencies and robust against errors in the control Hamiltonian. This feedback procedure is also shown to overcome spontaneous emission effects by stabilising states with high degree of entanglement.PACS numbers: 03.67. Mn,42.50.Lc,03.65.Yz Numerous applications of quantum information theory require the ability to produce entangled states and perform controlled operations on them. There have been many successful experiments in this direction [1,2,3,4,5] but despite the effort to screen the system against unwanted imperfections and interactions, entanglement degradation through uncontrolled coupling with the environment remains a major obstacle [6,7]. Even if experiments were to be performed under perfect conditions, fundamental factors such as spontaneous emission in atomic qubits [8] would persist, limiting the lifetime of entangled states and demanding efficient schemes to protect them.Recent experimental developments have enabled realtime monitoring and manipulation of individual quantum systems [9,10,11,12], suggesting that quantum feedback control [13,14,15,16], may emerge as a natural possible route to develop strategies to prepare entangled states and prevent their deterioration. Recent attempts have been made in this direction with proposals to control both continuous [17,18] and discrete [19,20,21] variable entanglement, using either Bayesian [16] or Markovian (direct) [14,15] feedback scheme. While in the latter strategy a simple feedback directly proportional to the detection signal is used, in the former, control depends on an estimate of the system state based on the information acquired from the measurement results. Although this can result in an improvement over the direct feedback scheme, it also comes at the cost of an increasing complexity in the experimental implementation due to the (challenging) need for a real time estimation of the quantum state.In this Letter we will show that a Markovian feedback scheme based on the continuous monitoring of quantum jumps, together with an appropriate choice for the feedback Hamiltonian, can lead to an improvement, in amount and robustness, of the steady state entanglement in a model of two driven and collectively damped qubits [21,22]. In the absence of spontaneous emission, a pure maximally entangled state is dynamically generated irrespective of detection inefficiencies. Furthermore, this strategy is also able to cope with spontaneous emission effects by stabilising highly entangled states.Our system consists of a pair of two-level atoms equally, and resonantly, coupled to a single cavity mode, with a coupling strength g. The atoms can spontaneously decay with rates γ 1 and γ 2 , and are simultaneously driven by a laser field (see Fig. 1). The cavity mode is damped, and in the limit where its decay rate κ is very large, it can be adiabatically eliminated leading to the following mast...