Related Servers

Dozens of tool pairs are functionally indistinguishable: apply_articulation vs set_articulation, split_stems vs separate_stems, create_break vs create_breakbeat, create_chord_progression vs create_chord_pads, create_arpeggio vs create_arpeggiated_progression, get_project_state vs get_full_project_state vs get_project_info vs get_project_metadata, and 30+ genre arrangement tools differing only in genre. An agent cannot reliably select between these without reading every description.

The dominant pattern is consistent verb_noun (create_*, set_*, get_*, list_*, delete_*) under a uniform mcp_opendaw_ prefix. However, there are notable deviations: bare nouns (transport, undo, redo, serialize, note_stats), inconsistent compound spelling (audio_unit vs audiounit in duplicate_audiounit vs delete_audio_unit), and mixed noun-verb forms (engine_sleep, ppqn_to_seconds).

515 tools is extreme by any standard — over 10x the high end of a typical well-scoped server. Even for a comprehensive DAW, the count is inflated by ~30 near-identical genre arrangement generators, at least 7 melody generation tools, 3 stem-splitting tools, 4 project-state tools, and dozens of micro-variations. The switch_phase mechanism acknowledges this bloat but doesn't reduce the total surface.

The DAW domain is covered end-to-end with no obvious dead ends: project lifecycle (create/load/save/reset), track/region/note editing, instruments, effects, sends/buses, automation, mixing, mastering, rendering/export (WAV/stems/MIDI/dawproject), audio analysis (BPM/key/LUFS/spectrum/stereo/dynamics/masking), transcription, stem separation, and AI-assisted composition. Every workflow from import to master has tooling.