Loading ToolBench
Maya plugin for the DCC Model Context Protocol (MCP) ecosystem — embeds a Streamable HTTP MCP server directly inside Maya
27 tools with consistent naming (verb_noun pattern) and structured schemas. Descriptions present but generic, most lack actionable context for LLM selection. Parameter descriptions are sparse or missing entirely. No output schemas documented. Error handling guidance absent. Tool composition is sound (single responsibility), but descriptions do not explain WHEN to use each tool or dependencies between them. Baseline: 194 chars avg description; these average ~80 chars with minimal context.
Create an Arnold AOV and wire it into the render options output list.
Connect a source into the next free layer of a comp stack.
Wire two comp nodes together.
Create a reverse / multiplyDivide / luminance / clamp / setRange node.
Create a file node that reads rendered imagery from disk.
Create an empty layeredTexture stack to composite layers into.
Create a renderSetup collection, or replace an existing collection's members.
Create a renderSetup render layer and optionally make it visible.
Parameter descriptions missing or minimal. Most parameters (e.g., 'plugin', 'stack', 'source') lack descriptions explaining what they control, expected format, or constraints. LLMs cannot infer parameter semantics from names alone.
Output schemas not documented. No tool specifies what fields are returned or their types. LLMs cannot plan downstream tool calls or extract required data without knowing response structure.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 65 | 2026-07-28+ | v2 |
Override an attribute for everything a renderSetup collection selects.
Explain why a plug-in is or is not usable: path, load state and problems.
Report the plug-in search path and resolve a plug-in name to a file.
List the AOVs wired into Arnold's render options.
List the layers currently connected into a comp stack.
List Maya plug-ins with their load state and metadata.
List renderSetup layers, the visible layer, and legacy renderLayer nodes.
Load a Maya plug-in, with an actionable error when it cannot be found.
Combine two or more sources into a single comp output.
Plan per-layer and per-AOV output paths for a compositing hand-off.
Read an existing project.json back
Return the rehydration payload an agent needs to restore scene, assets, active skills, tool groups and checkpoint IDs
Persist current Maya project state to .dcc-mcp/project.json
Pure-read snapshot of the current state
Disconnect and delete an Arnold AOV.
Disconnect one layer from a comp stack, keeping its source node.
Enable or disable an existing Arnold AOV without disconnecting it.
Switch the current render layer (renderSetup or legacy).
Unload a Maya plug-in, refusing when Maya reports it cannot be unloaded.
Tool descriptions are generic and lack actionable context. Most descriptions (40 - 60 chars) state WHAT the tool does but not WHEN to use it, dependencies, or prerequisites. Baseline is 194 chars; these are 50% shorter.
No error handling guidance. Tools do not document what errors can occur, how to recover, or what the LLM should do next. E.g., load_plugin says 'actionable error when it cannot be found' but no error schema or recovery steps are visible.
Destructive operations (remove_aov, remove_comp_layer) lack confirmation or dry-run support. Agents can delete resources without a safety gate, risking irreversible data loss.