No user-submitted related servers found.
Scored across 41 tools
Tool prefixes (scene_, agent_, workspace_, export_, artifact_) separate major domains, and most status/run pairs are distinguishable. A few near-duplicates remain—scene_synthetic_views vs scene_synthetic_view_image, scene_job_status vs workspace_job_status, and artifact_fetch vs artifact_fetch_splat/cloud—but the descriptions clarify intended use. Overall, the set is more coherent than most large toolkits.
Names consistently use snake_case and a prefix-based convention that encodes the domain (scene_, workspace_, agent_, export_, artifact_). However, not all tools follow a verb_noun pattern: scene_planes, scene_lod, scene_imported_objects, and scene_share_access are noun-phrase reads, and status vs build/run pairs vary slightly. The pattern is predictable once the prefixes are understood.
41 tools is excessive for an agent-facing MCP server, even for a broad scene-reconstruction pipeline. Many tools are thin status/trigger pairs (scene_lod/scene_lod_build, scene_imported_objects/scene_imported_objects_run) or wait/status variants that could be consolidated. The scope is broad, but the count creates real selection overhead.
The server covers the full ingest → reconstruct → inspect → annotate → export → artifact-fetch lifecycle with no dead ends in the core workflow. Gaps include scene deletion/metadata updates and direct object editing, but these are not required by the apparent purpose. Overall the surface is complete enough for real workflows.