Young massive clusters provide the opportunity to study the outcome of the star formation process and the early evolution of star clusters. In the first few million years, the (massive) stars dynamically interact, producing runaways and affecting the initial (binary) population. Observing and interpreting the dynamics of young massive clusters is key to our understanding of the star formation process and predicting the outcome of stellar evolution, for example the number of gravitational wave sources. We have studied NGC 6611 in the Eagle Nebula (M16), a young massive cluster hosting ~19 O stars. We used Gaia EDR3 data to determine the membership, age, cluster dynamics, and the kinematics of the massive stars including runaways. The membership analysis yields 137 members located at a mean distance of 1706+/-7pc. The colour - absolute magnitude diagram reveals a blue and a red population of pre-main-sequence stars, consistent with two distinct populations of stars. In line with earlier studies, the youngest (reddest) population has a mean extinction of A_V_=3.6+/-0.1mag and an age of 1.3+/-0.2Myr, while the older population of stars has a mean extinction of A_V_=2.0+/-0.1mag and an age of 7.5+/-0.4Myr. The latter population is more spatially extended than the younger generation of stars. We argue that most of the OB stars belong to the younger population. We identify eight runaways originating from the centre of NGC 6611, consistent with the dynamical ejection scenario. We have studied the kinematics of the O stars in detail and show that ~50% of the O stars have velocities comparable to or greater than the escape velocity. These O stars can be traced back to the centre of NGC 6611 with kinematic ages ranging from 0 to 2Myr. These results suggest that dynamical interactions played an important role in the early evolution of NGC 6611, which is surprising considering the relatively low current stellar density (0.1-1x10^3^M_{sun}_/pc^3^). Comparing our results to simulations of young massive clusters, the initial radius of 0.1-0.5pc (needed to produce the observed O star runaway fraction) is not consistent with that of NGC 6611. We propose a scenario where the O stars initially form in wide binaries or higher order systems and possibly harden through dynamical interactions.