MCP server wrapping the DazScriptServer HTTP API for DAZ Studio
44 tools with mixed quality. Naming is strong and verb-forward across all tools (daz_get_*, daz_set_*, daz_apply_*, etc.). Descriptions are present for all tools and average ~100-150 chars, which is appropriate. However, input schemas are partially visible, all tools show parameter names and types in the provided list, but JSON Schema completeness cannot be fully verified without inspecting source. The tools are organized logically across modules (animation, camera_light, morph, spatial, render, etc.), suggesting good composition. Critical gap: no visible output schema documentation, return types are not described, forcing LLMs to infer structure. Error handling approach uses broad exception-catching with a typed dispatch (handle_dazpy_error) per pylint rationale, which is appropriate for the domain but specifics are not visible in code excerpts. Security: no obvious credential exposure in parameter lists. Tool design follows single-responsibility principle well (e.g. daz_set_frame vs daz_set_frame_range are split). Batch operations present (daz_batch_select, daz_batch_visibility, daz_batch_set_properties), showing composition awareness. Moderate issue: daz_execute accepts arbitrary DazScript code, no validation or safety guardrails visible; this is a significant risk for prompt injection.
Apply a camera angle preset
Apply a composition rule to frame the scene
Apply a lighting preset to the scene
Select multiple nodes by label
Set properties on multiple nodes
Set visibility for multiple nodes
Browse assets in a content category
No documented output schemas for any of the 44 tools. LLMs cannot plan downstream tool calls or extract structured data without knowing what fields to expect. This is a critical gap for composition and chaining.
daz_execute tool accepts arbitrary DazScript code with no validation, sanitization, or safe-guard parameters (no dry-run mode). This enables prompt injection and unintended side effects. High security risk.
Tools accepting enumerated values (e.g., daz_set_emotion accepts 14 emotion names, daz_set_render_quality accepts quality levels, daz_frame_shot accepts shot_type) do not declare enums in their schemas. Descriptions include example values like 'happy, sad, angry, surprised', which LLMs may treat as exhaustive lists or hallucinate variations. Enums should be formal JSON Schema constraints.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 58 | 2026-07-28+ | v2 |
Calculate distance between two nodes
Check if two nodes' bounding boxes overlap
Execute arbitrary DazScript code
Find nodes within a specified distance of a target node
Frame the camera to focus on a specific node
Frame a cinematic shot on the current camera
Get animation frame range, current frame, and playback state
Get the bounding box of a node
Get information about a node
Get the scene hierarchy starting from a node
Get metadata about a property
Get current render settings
Get status of a render request
Get the spatial layout of the scene
Get spatial relationship between two nodes
Get the world position of a node
List properties of a node
List content categories
List saved scene checkpoints
List child nodes of a parent node
List available morphs for a figure
List all active and completed render requests
Load a camera preset from the content library
Orbit the camera around a target node
Restore a previously saved scene state/checkpoint
Save the current camera view as a preset
Save the current scene state as a checkpoint
Get basic scene information
Search for morphs by name or category
Set the active camera in the scene
Set an emotion preset on a figure
Set the current animation frame
Set the animation frame range
Set render quality level
Get server status
Validate the current scene for common issues
Validate DazScript code for syntax and common issues
No visible error handling guidance for any tools. When a tool fails (e.g., node not found, asset doesn't exist), there is no documentation on what error message is returned or how the LLM should recover (retry, try alternative tool, ask user). This prevents graceful degradation and multi-step planning.
Batch tools (daz_batch_select, daz_batch_visibility, daz_batch_set_properties) accept arrays but do not document per-item success/failure reporting. If one item in a batch fails, does the entire batch fail or does it return a mixed result? Unclear specification invites misuse.
Many list/browse tools (daz_list_categories, daz_list_checkpoints, daz_list_children, daz_list_morphs, daz_list_requests) do not specify pagination parameters (limit, offset, page_size) or total count in response. Large results could blow context window without pagination.
Tool descriptions are generic and do not explain WHEN or WHY to use a tool. For example, daz_get_animation_info is described as 'Get animation frame range, current frame, and playback state' but does not explain when an LLM should call this vs. other animation tools or what the use case is. Descriptions should include dependency hints and context.