No user-submitted related servers found.
Scored across 156 tools
With 156 tools there is heavy overlap: get_ai_ml_package_inventory / get_ai_stack_summary / get_unity_ai_stack all report AI package stacks, list_ml_model_assets / list_inference_models / list_ml_agents_training_configs all scan model files, and get_localization_stack vs get_localization_tables are near-duplicates. Dozens of narrow list_* asset enumerators (list_assets_by_extension, list_scriptable_objects, list_sprite_assets, list_tmp_fonts) blur into one another, so an agent must guess which one to pick.
Nearly everything follows a predictable snake_case verb_noun pattern (list_*, get_*, find_*, search_*, scan_*, detect_*), which is readable and largely consistent. Minor deviations like bridge_live_fetch (noun_verb) and overlapping verb choices (get_package_dependency_graph vs get_assembly_dependency_graph vs detect_assembly_cycles) prevent a perfect score.
156 tools is roughly ten times a well-scoped surface and far beyond what an agent can reliably select from. Many tools are extremely narrow single-purpose audits (get_audio_loudness_risk, get_timeline_balance, list_lottie_assets) that would be better folded into a few parameterized tools.
Coverage of read-only project inspection is genuinely exhaustive—settings, assets, packages, build, testing, and third-party SDK detection are all present. However the surface is entirely read/analysis: there is no create, update, or delete operation for any resource, and no Editor mutation or build-execution path, leaving a notable gap for an MCP server meant to act on a Unity project.