How MusEdLab works
The AI powering your tools, how your data is handled, and what you should know about working with AI in your teaching practice.
Powered by Claude AI
Every AI-powered tool in MusEdLab is built on Claude, the AI assistant made by Anthropic. Claude is one of the most capable large language models available, and it's specifically designed to be helpful, accurate, and safe.
We chose Claude because of its strong reasoning abilities, its deep knowledge base spanning music theory, history, and education, and because Anthropic has made responsible AI development a core part of their mission — something we care about deeply as educators.
Matt Woodward, MusEdLab co-founder, is certified in Anthropic's AI Fluency & Foundations program and has trained thousands of educators on practical AI use across 38 states.
How AI reads your scores
When you upload a PDF of sheet music, here's what happens under the hood:
The AI reads notation from the visual content of the PDF. Handwritten scores, very low-resolution scans, or heavily marked-up copies may produce less accurate results. Clear, high-contrast printed scores work best.
Why your ensemble context matters so much
AI language models like Claude don't automatically know who you are, who your students are, or what you're trying to accomplish. When you provide detailed ensemble context, you're essentially briefing the AI so it can tailor every output specifically to your situation.
The same piece of music can produce very different — and much more useful — reports depending on whether you describe:
- A beginning 5th-grade band preparing for their first concert
- An advanced high school choir preparing for a state adjudication
- A college chamber ensemble performing for a conducting seminar
The AI adjusts vocabulary, difficulty assumptions, recommended strategies, and even tone based on your description. See Writing Great Ensemble Context for detailed guidance.
Data privacy
We take privacy seriously. Here's exactly what happens with your data:
What we store
- Your Google account email and name (for login and account management)
- Your usage count (to enforce free tier limits)
- Your subscription status (if you're a Pro member)
What we do NOT store
- Your uploaded PDFs — deleted immediately after processing
- The text of your analyses or generated reports
- Your ensemble context descriptions
- Your follow-up chat conversations
Analysis results are stored only in your browser session. When you close the tab or start a new analysis, the previous results are gone from our systems. This means you may want to print or export important reports before closing.
Anthropic's data handling
Your PDFs and prompts are processed through the Anthropic API. Anthropic does not use API inputs and outputs to train their models by default. For more information, see Anthropic's Privacy Policy.
Copyright & uploads
Copyright law applies to sheet music just as it does to books and recordings. Before uploading any music to MusEdLab, you must ensure:
- You own a legal copy of the sheet music, OR
- The music is in the public domain (generally, works published before 1928 in the United States)
You are solely responsible for ensuring your uploads comply with copyright law. MusEdLab is designed for rehearsal preparation using music you legally own — not for digitizing or distributing copyrighted scores.
Public domain music
If you're looking for public domain scores to use freely, excellent sources include:
- IMSLP / Petrucci Music Library — the largest repository of public domain sheet music
- Choral Public Domain Library (CPDL) — specifically for choral music
- MuseScore Free Use
AI limitations to understand
AI is a powerful tool — but it's not perfect, and understanding its limitations will help you use MusEdLab more effectively.
The AI doesn't know your students personally
No matter how good the AI is, it has never heard your choir sing or watched your band rehearse. Every recommendation is based on what you tell it. Your professional judgment as an educator should always be the final filter.
The AI's knowledge has a cutoff
Claude's training data has a knowledge cutoff date, which means it may not know about very recently published music or newly premiered works. For newer repertoire, results will be less detailed.
Notation reading is impressive but not perfect
Claude reads scores visually — the same way a human would scan a page. Unusual notation, non-standard engraving, or poor scan quality can occasionally lead to misidentified passages or missed markings. Always cross-reference with the score itself.
Generated content should be reviewed
Program notes, lesson plans, and practice guides are excellent first drafts — but they should be reviewed, personalized, and edited by you before being shared with students, parents, or administrators. AI output is a starting point, not a finished product.
Our tech stack
For the technically curious, here's what MusEdLab is built with:
- Claude AI (Anthropic) — the AI powering all tool outputs
- Python / Flask — our backend web framework
- Google OAuth 2.0 — secure sign-in without passwords
- Replit — our development and hosting platform
- Stripe — payment processing for Pro subscriptions
We're a two-person team building lean and deliberately. Have feedback on the tech or the product? We'd love to hear it.