Pixal3D-MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| load_pipelineA | Load the Pixal3D pipeline and all required models into GPU memory. This must be called before running inference. |
| unload_pipelineA | Unload the Pixal3D pipeline and free GPU memory. |
| run_inferenceC | Run Pixal3D image-to-3D inference. CLI-style args: --image --output --seed --max_num_tokens |
| get_memory_statusB | Get current GPU memory usage and pipeline status. |
| cleanup_memoryA | Clean up GPU memory by running garbage collection and clearing CUDA cache. |
| load_conda_envA | Load environment variables from a conda environment. Use this when the MCP server is launched from a desktop app (like Zed) without the conda environment pre-activated. |
| get_conda_env_statusB | Get the current conda environment status and configuration. |
| list_conda_envsA | List all available conda environments. Use this to discover what conda environments you can load with load_conda_env. |
| run_inference_sequenceB | Run Pixal3D inference on an image sequence. Supports Maya, After Effects, and Fusion filename conventions. Auto-detects frame padding, sequence pattern, and frame range. |
| preview_imagesA | Generate thumbnail previews for images without running inference. Fast, non-blocking operation that returns instant visual feedback. Supports sequences, IFL files, and glob patterns. |
| preview_modelB | Render GLB models to preview PNG images using 3d-to-image. Fast, non-blocking. Supports single files, sequences, and IFL files. |
| preview_model_sequenceC | Render a GLB model sequence to preview PNG images with optional frame range filtering. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 12 tools
Most tools have clearly distinct purposes (pipeline load/unload, inference, memory status vs cleanup, conda env management). The main overlap risk is between run_inference/run_inference_sequence and preview_model/preview_model_sequence, where the singular vs sequence variants could be confused, though the descriptions distinguish them adequately.
All 12 tools use a consistent snake_case verb_noun pattern (load_pipeline, run_inference, get_memory_status, preview_model_sequence). Verb prefixes (load/unload/run/get/cleanup/list/preview) are used predictably throughout.
12 tools is well-scoped for a 3D inference server, with distinct functional clusters (pipeline lifecycle, inference, memory, conda env, previews). Each tool appears to earn its place without redundancy.
The surface covers pipeline lifecycle, inference, memory management, conda environment handling, and preview rendering for both images and models. Minor gaps exist (e.g., no explicit batch/config inspection or pipeline config tools), but core workflows are well covered.