mcp

Most tools have clear, distinct purposes with detailed descriptions, but a few overlap: get_study_results and export_study_results return identical data with only intent differing, and move_block vs reorder_blocks both handle positioning. The single/bulk pairs (add_block vs bulk_add_blocks) are distinguishable by cardinality, so overall ambiguity is low.

Tool names predominantly follow a verb_noun snake_case pattern (e.g., list_studies, create_study, delete_block), with bulk_ prefixes and get_/list_ conventions. Minor deviations include show_trace instead of get_trace, and prepositions in build_from_plan and move_to_project, but these are predictable and readable.

With 38 tools, the server is well beyond the 25-tool threshold for a typical MCP server. Although each tool has a distinct role and the domain is broad, the count feels heavy, especially with single/bulk variants that could be consolidated (e.g., add_block vs bulk_add_blocks, update_block_config vs bulk_update_blocks).

The tool surface covers the entire study lifecycle: planning (scaffold_plan, save_plan), building (create_study, add_block, bulk_add_blocks, update_block_config, delete_block, reorder_blocks), launching (launch_study, pause_spawn, stop_spawn, archive_spawn), tracking (attach_tracker, verify_tracker, get_tracker_events), results (get_study_results, export_study_results), and analysis (list_analyses, get_analysis). No obvious gaps or dead ends were identified.