The camera streams are HLS via AVFoundation – Apple doesn't expose raw frames, so no programmatic capture there. But motion detection events do flow through the SSE endpoint (/events), so you can use those to trigger webhooks or automations. For HA cameras there's also a snapshot endpoint you can hit on demand, e.g. when motion fires.

SSE is in the upcoming 2.0.0 which is currently on TestFlight – happy to add you if you want to try it out.