Loading ToolBench
🎬 The FIRST MCP server for Kling AI video generation! Generate, extend, lip-sync AI videos, virtual try-on, and manage your account directly from Claude.
The server defines 11 tools with explicit schemas and descriptions. Most tools have reasonable parameter documentation and appropriate enums. However, there are consistent gaps: output schemas are undocumented across all tools (pattern:tool violation), many parameter descriptions lack actionable constraints (e.g., no guidance on image URL format, video URL accessibility), and error handling is absent (no recovery guidance, no categorization of retryable vs fatal errors). Tool naming is verb-first (good), but some descriptions are generic and lack WHEN-to-use guidance. Parameter validation rules are not described (e.g., what makes a valid image_url? Must it be publicly accessible?). The camera_control nested object in generate_video is well-structured but represents edge case complexity that could be clearer. Overall: functional definitions that would work with an agent, but below production grade due to missing output schemas and error guidance.
Check the status of a video generation task
Create a lip-sync video by synchronizing mouth movements with audio. Supports both text-to-speech (TTS) with various voice options or custom audio upload. The original video must contain a clear, steady human face with visible mouth. Works with real, 3D, or 2D human characters (not animals). Video length limited to 10 seconds.
Apply visual effects to images to create animated videos with specific effects like hug, kiss, heart gesture, squish, expansion, and bloom
Download a generated video to local disk
Extend a video by 4-5 seconds using Kling AI. This feature allows you to continue a video beyond its original ending, generating new content that seamlessly follows from the last frame. Perfect for creating longer sequences or adding additional scenes to existing videos.
Generate an image from a text prompt using Kling AI
No output schemas documented for any tool. Pattern:tool and pattern:response-shaper require structured return type documentation so LLMs know what fields to expect and can chain tool calls. Without documented outputs, agents cannot reliably extract task_id from generate_video to pass to check_video_status.
No error handling or recovery guidance. When generate_video fails (e.g., API rate limit, invalid image_url, authentication failure), the implementation provides no actionable guidance for the LLM. Pattern:recovery-guide and pattern:error-classification are absent.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-21 | C | 60 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 28 | - | v1 |
Generate a video from an image using Kling AI
Generate a video from text prompt using Kling AI
Get account balance and resource package information
List all video generation tasks with optional filtering by status and date range
Create a virtual try-on video showing how clothing items would look on a person. Perfect for fashion e-commerce and styling applications.
Parameter descriptions lack actionable constraints. image_url, image_tail_url, video_url, audio_url, person_image_url, and cloth_image_urls parameters state 'URL' but do not specify: must URLs be publicly accessible? Are signed URLs supported? What image formats? What file size limits? These details force the agent to guess or make unnecessary validation calls.
Pagination in list_tasks lacks explicit total count and next_cursor. Pattern:paginated-result requires that list tools return a total count so agents can determine if more pages exist. Without it, agents cannot reliably iterate through all tasks.
Descriptions for most generation tools are generic. 'Generate a video from text prompt using Kling AI' and 'Generate an image from a text prompt using Kling AI' lack WHEN-to-use guidance. Pattern:tool-description requires context for selection. E.g., when should an agent choose generate_video vs generate_image_to_video?
download_video modifies local filesystem state (writes to ~/Desktop or custom path) but has no dry-run or confirmation mechanism. Pattern:confirmation-request suggests gating side effects. An agent could overwrite existing files without user awareness.