Related Servers

Most tools target distinct operations (lip sync, talking photo, motion transfer, extend, task polling), but the generation surface is fragmented across six near-siblings: kling_generate_video, kling_generate_video_from_image, kling_generate_turbo_video, kling_generate_storyboard, kling_generate_with_assets, and kling_goods_studio. The boundary between plain generation and model-specific variants (turbo/storyboard/assets) is only weakly explained. kling_list_models and kling_list_actions also overlap as discovery/reference tools.

Nearly all names follow a kling_verb_noun convention in consistent snake_case with a uniform prefix (list_models, get_task, generate_video, extend_video). The exceptions are kling_goods_studio and kling_video_commerce, which are noun phrases rather than verb-led, plus the redundant 'kling_' prefix on every tool. These are minor deviations rather than a pattern break.

At 15 tools the server sits at the top of the well-scoped band, which is defensible for an API with multiple model tiers and specialized modes. A few tools (generate_turbo_video, generate_storyboard, generate_with_assets) are thin wrappers around model variants that arguably could be parameters. Slight over-fragmentation, but each tool maps to a real distinct endpoint.

Coverage is strong across the generation lifecycle: text/image/motion/storyboard/turbo generation, lip sync, talking photo, extend, plus single and batch task polling. Gaps are minor but real: no cancel/delete task, no list-tasks/history, and no account credit or balance query, which matters given the tool descriptions mention credit costs. Agents can work around these but not without friction.