We consider a broad class of systems of nonlinear integro-differential equations posed on the real line that arise as Euler-Lagrange equations to energies involving nonlinear nonlocal interactions. Although these equations are not readily cast as dynamical systems, we develop a calculus that yields a natural Hamiltonian formalism. In particular, we formulate Noether's theorem in this context, identify a degenerate symplectic structure, and derive Hamiltonian differential equations on finite-dimensional center manifolds when those exist. Our formalism yields new natural conserved quantities. For Euler-Lagrange equations arising as traveling-wave equations in gradient flows, we identify Lyapunov functions. We provide several applications to pattern-forming systems including neural field and phase separation problems.Differential equations of gradient form u t "´∇Epuq or of Hamiltonian form u t " J∇Epuq arise throughout mathematical modeling from maximal energy dissipation first principles, or from least action principles, respectively. Mathematically, the energy E is simply a function defined over a finite or infinite-dimensional space. The associated gradient and Hamiltonian flows are differential equations with equilibria given by the critical points of the energy. Energies over infinite-dimensional spaces are usually defined on function spaces through integrals of nonlinear functions of state variables and their derivatives. Critical points then solve Euler-Lagrange equations, commonly of elliptic type. Dependence of the energy on derivatives of the state variables encodes local interactions, often derived in various types of continuum limits. We are interested here in cases where this continuum limit retains nonlocal interaction terms. More specifically, we are interested in the somewhat specific class of energies E that contain nonlocal interaction term of the formmodeling phenomena such as long-range interactions of agents in social interaction, between particles through nonlocal force fields, or of neurons, labeled in a feature space x through synaptic connections. In all those cases, the convolution structure embodies the modeling assumption of translational invariance of physical space. We are