Text-to-MIDI plug-in MIDI Agent discloses no audit trail, insurance or runtime limits of its own

MIDI Agent, a text-to-MIDI plug-in sold to music producers as a one-time "$49 /lifetime" license, discloses no audit trail, insurance, indemnity or usage limits of its own, according to an AI Dispatch index review completed this week.

The plug-in is "available as VST3, AU, AAX, and Standalone," per the vendor's site, and "uses Large Language Models (LLMs) like ChatGPT and Claude to create MIDI data from natural language descriptions," letting users "generate melodies, chords, drums, and basslines from prompts with ChatGPT, Claude, and Gemini." Users connect their own ChatGPT account or supply their own provider API keys; any cap on what the tool can do comes from that outside account rather than from MIDI Agent itself, which states that "available models and usage limits depend on your ChatGPT plan and workspace."

On data handling, the vendor states "user interactions, input files, or output files are not used for training AI models by MIDI Agent." It also discloses a limit on what the tool can produce: because its "models are not trained on copyrighted music material," it warns that "using an artist's name as a reference is very unlikely to yield the desired results or may be denied entirely" — a disclaimer on output quality, not a guarantee of it.

An AI Dispatch index review scored MIDI Agent zero out of three on audit trail, insurance/indemnity, outcome-based settlement and runtime governance. It scored 1 out of 3 on compliance, based solely on the training-data disclosure above — no certification, auditor or compliance schedule is published. No insurance carrier, coverage limit or customer indemnity appears anywhere on the vendor's site, and the license fee is owed regardless of what the tool produces.