Skip to main content

What is MRS?

MRS (Music Representation Syntax) is a format for encoding musical scores optimized for AI-assisted composition:
  • MRS-S: An S-expression-based format for complete, archival-quality scores
  • Working Set Envelopes: Scoped MRS-S fragments with metadata for bounded agent operations
The key insight: Agents don’t need a different format—they need a different scope. MRS uses a single syntax throughout; Working Set Envelopes simply extract bounded fragments with explicit boundaries. MRS is designed first and foremost to unlock AI & agent-driven development workflows on full-scope musical scores—where tools (and machine agents) can reliably read, modify, validate, and merge changes without needing the entire score in one context window. While the syntax is intentionally human-auditable, it is not primarily optimized for manual entry.

How It Works

Full Score (MRS-S)                    Working Set Envelope
┌─────────────────────────┐           ┌─────────────────────────┐
│ 90 parts × 1000 measures│  extract  │ Scoped MRS-S fragment   │
│ ~6 MB                   │ ────────► │ + scope metadata        │
└─────────────────────────┘           └─────────────────────────┘

                                                  │ agent edits

                                      ┌─────────────────────────┐
                                      │ Agent returns:          │
                                      │ Valid MRS-S fragment    │
                                      └─────────────────────────┘

┌─────────────────────────┐           replace +   │
│ Updated full score      │ ◄──────── validate ───┘
└─────────────────────────┘

How to use these docs

Normative source

This site is an organized presentation of the RFC document in /MRS-Specification-RFC. Where wording matters, treat the RFC as the normative reference.