The classification of primate gestures is traditionally achieved through top-down approaches, i.e. the categorization of gesture types based on the assignment of each gesture to a pre-described category. Being based on researchers’ intuition, these classifications may lack consistency in the applied criteria. For example, the gesture poke is characterized by its stretched fingers handshape, while other gestures with the same basic movement are all classified as touch, regardless of whether the hand, knuckle, or foot is used. Despite these limitations being acknowledged and occasionally addressed, there is still to date no coding scheme that formally describes primate gestures without relying on pre-established categories. This study aimed to fill this gap by developing an annotation system that describes chimpanzee gestures primarily through formal parameters, and only subsequently applies them to the identification of gesture types. To ensure we focused on formal features that may be relevant for meaning, we applied and adapted to chimpanzees an existing linguistic annotation system for co-speech gestures in humans, known as LASG (Bressem et al., 2013). The resulting coding scheme, ChimpLASG, comprises 14 formal parameters: it firstly subdivides gestures into phases, identifying their smallest unit, the stroke; then, of each stroke, it describes the shape (hand-wrist-elbow configuration, palm and forearm orientation, and position), the form of movement (movement descriptor, type, direction, trajectory, and quality), and type of touch (touch quality, body part touching, body part touched). Currently, ChimpLASG is fully developed and undergoing its first application for video coding in ELAN on a dataset of semi-wild chimpanzees in Chimfunshi, Zambia. Form-based approaches like ChimpLASG overcome current biases in primate gesture classifications and enable further investigations on gestural units, compositionality, and co-variation of gesture formal parameters with demographic, social, and ecological factors.