Loading ToolBench
Image-to-3D mesh generation application with neural rendering, texture baking, and multi-format export. Supports NVIDIA CUDA, DirectML, and CPU inference.
MeshForge 3D has 14 tools with basic naming and descriptions, but significant gaps in schema completeness, parameter documentation, and error handling. Most tools lack detailed input parameter descriptions beyond type declarations. Output schemas are not documented. Error handling is minimal, no recovery guidance or actionable error messages visible. Tool names follow verb_noun convention (get_*, set_*, generate_*, etc.), which is good, but descriptions are often too brief (10-50 chars) to guide LLM selection effectively. The codebase shows Eel-based UI integration but no evidence of MCP protocol implementation, transport is unknown, likely STDIO or local socket only.
Activate a MESHF-XXXX-... key purchased from the seller.
Bake neural texture onto mesh UV atlas using xatlas and neural rendering.
Cancel ongoing 3D generation process.
Check if model checkpoint is downloaded and ready.
Download TripoSR model checkpoint from Hugging Face Hub.
Export generated mesh to specified format (glb, obj, stl, ply).
Main image-to-3D generation pipeline.
Get current generation progress and status.
Output schemas not documented. Tools return data but LLMs cannot predict field names, types, or structure. Breaks downstream tool chaining and forces agents to guess at response format.
Descriptions too brief (10-50 chars). Lack context for LLM selection. E.g., 'Return hardware capabilities and model status for the frontend' (57 chars) does not explain WHEN to call it or what 'hardware capabilities' means. Baseline: 194 chars average.
Parameter descriptions missing or minimal. 'options_json' in generate_3d has a description but no validation rules, format constraints, or examples of valid JSON structure. LLMs cannot construct valid payloads without explicit format guidance.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 54 | 2026-07-28+ | v2 |
Current license tier for UI gating of Pro features.
Get geometry statistics for a generated mesh.
Return available sample images for 1-click test.
Return hardware capabilities and model status for the frontend.
Switch execution device (e.g. 'cuda:0' or 'cpu').
Generate concept image from text prompt using Stable Diffusion.
No error handling or recovery guidance visible. Tools do not document what errors can occur, whether they are retryable, or what the LLM should do next. E.g., download_model may fail if Hugging Face is unreachable, but no guidance provided.
No idempotency or confirmation for destructive operations. cancel_generation and export_mesh modify state but lack dry-run, confirmation, or idempotency hints. Agents may accidentally cancel mid-generation or overwrite exports.