Loading ToolBench
MCP server for kie.ai — 60+ image models, 85+ video models, 20+ audio tools with deep model intelligence
kie-mcp provides 18 tools with complete input schemas and descriptions, but quality is uneven. Naming is verb-forward and clear (generate_image, extend_music, list_models). Descriptions range 50 - 200 chars, meeting baseline. However, output schemas are undocumented, LLMs cannot predict return structure for chaining. Error handling is minimal; no recovery guidance or actionable error messages visible. Parameters lack enums for constrained values (e.g., model identifiers, voice styles). Security: API key injected via env (good), but file path parameters accept absolute paths without traversal validation, risking directory escape. Composition is strong, tools are single-purpose and chainable (e.g., generate_image → image_to_image). Per-tool scores average 62; no tool exceeds 75 due to missing output schemas and error guidance.
Generate an MP4 music video visualization from a Suno track for social media sharing
Extend an approved Suno music track from a timestamp to create length variants (15s/30s/60s) of the same identity
Generate multi-speaker dialogue using ElevenLabs Text-to-Dialogue V3 with distinct voices per character
Generate voiceover and narration using Google Gemini TTS with 30 named voices, style control, and tone tags
Generate images from text prompts using various AI models (GPT-4o, Flux, DALL-E, Qwen, Ideogram, etc.)
Generate song lyrics using Suno, with style and mood control (max 200 characters)
Output schemas undocumented. No tool describes what fields are returned, preventing LLMs from planning downstream calls or extracting results. E.g., generate_image returns what? URL? File path? Metadata?
No error handling guidance. Tools do not document what errors can occur, whether they are retryable, or what the LLM should do next. E.g., if generate_image fails due to invalid model, should the LLM retry, call list_models, or ask the user?
Model and voice parameters lack enums. 'model' accepts free-form strings (e.g., 'gpt4o', 'flux-kontext-pro') without constraint. LLMs will hallucinate invalid model names. Should enumerate valid options or reference list_models.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2026-07-28+ | v2 |
Generate music and audio tracks using Suno V5 with lyrics, structure hints, and style control
Generate sound effects and sonic logos using Suno V5 text-to-SFX (UI clicks, stings, whooshes, ambiences)
Generate text-to-speech using ElevenLabs Turbo 2.5 or Multilingual V2 for voiceovers and narration
Generate videos from text prompts or images using video AI models (Veo, Runway, Kling, Sora, etc.)
Get pricing information for a specific model or all models, with cost estimates and credit calculations
Edit images using Grok Imagine Image 2.0 with region-based editing and segment mapping
Generate a segment map of an image showing named regions for targeted editing (FREE operation)
Transform or edit images using AI models with reference images and prompts
List all available image, video, and audio models with pricing, capabilities, and parameter schemas
Get expert guidance for a vertical domain (architecture, advertising, game assets, etc.) with intake questions and model routing
Split an approved audio track into stems (vocals, instruments, etc.) for mixing and ducking
Upload a local file (image, video, audio) to kie.ai for use as a reference or input in generation tasks
Path traversal risk. download_dir and file_path parameters accept absolute paths without validation. Code calls isAbsolute() but does not prevent '../' sequences or symlink escapes. Malicious agents could write outside intended directory.
Pagination missing. list_models and get_pricing do not document limit/offset or total count. If model registry grows large, responses could exceed context window without pagination support.