Loading ToolBench
Focused MCP-compatible HTTP bridge and stdio transport for Seedance 2 video generation via MuAPI.
Server defines 6 tools with complete JSON schemas and descriptions. Naming follows verb_noun convention (seedance_2_*, muapi_*). Schemas are well-structured with proper types, enums, and constraints. However, descriptions are generic and lack LLM-optimized guidance on WHEN to use each tool vs alternatives. Parameter descriptions are present but minimal (10-30 chars). No error handling guidance, no output schema documentation, no tool annotations. Missing idempotency hints for async operations.
Read the current MuAPI credit balance.
Poll a MuAPI prediction until it completes or fails.
Generate a Seedance 2 transition from a first frame, optionally to a last frame.
Animate one or more reference images with Seedance 2. One image anchors the start frame; multiple images enable reference-driven generation.
Generate a Seedance 2 video guided by image, video, and audio references.
Generate a Seedance 2 video from a text prompt. Use quality=fast for the fast route.
Tool descriptions lack LLM-optimized guidance. Descriptions are 30-80 chars but do not explain WHEN to use each tool vs alternatives (e.g., text_to_video vs image_to_video vs omni_reference). LLMs cannot infer selection criteria from generic descriptions.
No output schema documentation. Tools return async request_id but no schema for the final video result, metadata, or error structure. LLMs cannot plan downstream steps or extract relevant fields.
No error handling guidance. Tools can fail (invalid prompt, API errors, quota exceeded) but error responses do not tell LLM what to do next. No recovery hints or error categorization (retryable vs user-fixable).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 64 | 2025-06-18+ | v2 |
Parameter descriptions are minimal (10-30 chars). 'Detailed scene, subject, camera, lighting, and motion description' for prompt is helpful, but 'Reference image URLs' lacks format guidance (JPEG/PNG/WebP?), size limits, or resolution requirements.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Seedance generation tools are destructive (consume credits) and async (not idempotent if retried). LLMs need these hints to plan safely.