We introduce neutral excitation density-functional theory (XDFT), a computationally light, generally applicable, first-principles technique for calculating neutral electronic excitations. The concept is to generalise constrained density functional theory to free it from any assumptions about the spatial confinement of electrons and holes, but to maintain all the advantages of a variational method. The task of calculating the lowest excited state of a given symmetry is thereby simplified to one of performing a simple, low-cost sequence of coupled DFT calculations. We demonstrate the efficacy of the method by calculating the lowest single-particle singlet and triplet excitation energies in the wellknown Thiel molecular test set, with results which are in good agreement with linear-response timedependent density functional theory (LR-TDDFT). Furthermore, we show that XDFT can successfully capture two-electron excitations, in principle, offering a flexible approach to target specific effects beyond state-of-the-art adiabatic-kernel LR-TDDFT. Overall the method makes optical gaps and electron-hole binding energies readily accessible at a computational cost and scaling comparable to that of standard density functional theory. Owing to its multiple qualities beneficial to high-throughput studies where the optical gap is of particular interest; namely broad applicability, low computational demand, and ease of implementation and automation, XDFT presents as a viable candidate for research within materials discovery and informatics frameworks. The first-principles calculation of the excited-state energies of quantum systems holds crucial importance for the study of solar cells 1 , organic light emitting diodes 2 , and chromophores in biological systems 3 , to name but a few. With some exceptions, density-functional theory (DFT), the primary ab initio workhorse for computing ground state properties 4,5 , typically falls short on such tasks, although efforts are underway to extend the foundation of DFT to excited states 6-11. The most commonly used first-principles method for calculating excitation energies, at least of finite systems, is linear-response time-dependent density functional theory (LR-TDDFT) 12-14. However, LR-TDDFT has two significant limitations: its computational cost, which severely limits the size of the systems that it can be used to investigate 15 , and its inability to treat double (two-electron) or higher-order excitations within adiabatic approximations to the exchange-correlation (XC) kernel 16-18. Such multi-particle excitations and indeed inter-conversions between excitation types are of considerable interest, not alone on fundamental grounds, but for example in the computational development of photovoltaic materials that intrinsically overcome the Shockley-Queisser limit. Often, it is the lowest excitation energy of a given symmetry that is of principal interest, as this can be used to calculate the electron-hole binding energy. In such cases, state-of-the-art methods that rely on the coupling of ...