Scored across 39 tools
Most tools target distinct resources or actions, but potential overlaps exist in runtime session tools (run_project vs attach_project, detach_project vs stop_project) and profiling tools (start_profiler/profile_project/stop_profiler) that could cause misselection without careful reading. Descriptions are detailed enough to disambiguate in most cases.
Names are predominantly snake_case with a verb_noun pattern (get_project_info, add_node, connect_signal), with minor deviations like batch_scene_operations (noun phrase), validate (verb only), and start_profiler/stop_profiler vs profile_project (inconsistent profiler naming). Overall consistent and readable.
39 tools is well above the 15-tool upper bound for a well-scoped set. While the server covers a broad domain (project discovery, scene editing, runtime control, profiling, autoloads, validation), the sheer number increases cognitive load and overlaps; splitting or consolidation would help.
The surface covers core CRUD for nodes, signals, autoloads, and scenes, plus runtime execution, profiling, and validation. Minor gaps exist for file operations (delete/rename scenes or assets) and higher-level editing (animations, tilemaps), but agents can work around most with existing tools.