Read-only MCP server for searching C Dance AI Seedance 2.0 prompts and building structured video prompt briefs.
Four well-named tools with complete Zod schemas and clear descriptions. All tools are read-only with proper annotations. Naming follows verb_noun pattern (search_prompts, get_prompt, list_workflows, build_video_prompt_brief). Descriptions are substantive (100-200 chars), explaining WHAT and WHEN. Input schemas are fully typed with constraints (min/max lengths, enums, defaults). However, output schemas are not formally documented, responses are text-wrapped without explicit field definitions. Error handling is minimal (only get_prompt has try-catch). Parameter descriptions are present but could be more prescriptive about format/constraints.
Turn a rough video idea into a structured, deterministic Seedance prompt brief with motion, camera, continuity, audio, and revision guidance. This does not generate video.
Get one complete sanitized Seedance 2.0 prompt and its media by slug. Use a slug returned by search_prompts.
List practical Seedance 2.0 creation workflows with search queries for finding relevant public examples.
Search the public C Dance AI Seedance 2.0 prompt gallery. Returns concise previews, video thumbnails, and canonical source pages. Read-only and free.
Output schemas not formally documented. Responses are text-wrapped without explicit field definitions. LLMs cannot reliably extract structured data or plan downstream calls.
Minimal error handling. Only get_prompt has try-catch; others assume success. No recovery guidance for LLMs (e.g., 'If search fails, try a simpler query').
Parameter descriptions lack format/constraint details. E.g., 'language' says 'Language filter' but doesn't explain valid values or what 'all' means. 'duration' and 'aspectRatio' accept free-form strings without format guidance.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 65 | 2026-07-28+ | v2 |
No pagination support in search_prompts. Limit defaults to 8 and caps at 20, but no cursor/offset or total count returned. Large result sets could exhaust context.
build_video_prompt_brief returns unstructured text via buildVideoPromptBrief() function. No visible schema for motion/camera/audio/revision sections. LLM cannot reliably parse structured prompt brief.