Iris Tailor

The expert sewing friend you don't have.

Every fitted sewing project starts with a muslin — a test garment in cheap fabric, sewn to check the fit before cutting into good material. Then comes the moment a beginner gives up and an intermediate googles for an hour: standing in front of a mirror, trying to figure out why it pulls there, gapes here, bunches at the back. An experienced friend would just glance over and say "you need a swayback adjustment."

Iris Tailor is that friend. Upload your pattern, your measurements, and a photo of yourself in the muslin. Claude Opus 4.7 — orchestrated as specialist agents via Managed Agents — diagnoses fit issues from the photo, cascades corrections through the pattern, and animates each change so you learn pattern-making while the work happens.

Built solo for Anthropic's "Built with Opus 4.7" hackathon, April 2026.

Status

🚧 In development. Hackathon deadline: April 27, 2026.

Getting started

Prerequisites

• Python 3.11+
• Node.js 20+
• uv for Python
• pnpm for Node
• SMPL model files (register at smpl.is.tue.mpg.de, place in assets/smpl_models/)

Install

# Backend
cd backend
uv sync

# Frontend
cd ../frontend
pnpm install

Environment

cp .env.example .env.local
# Edit .env.local and add your API keys

Run

# Backend (in one terminal)
cd backend && uv run uvicorn main:app --reload

# Frontend (in another)
cd frontend && pnpm dev

Open http://localhost:3000.

Architecture

See docs/v2-plan.md for the full plan, week schedule, and architectural decisions.

In short:

• Claude Opus 4.7 via Managed Agents does the reasoning (fit diagnosis, cascade orchestration).
• Deterministic Python handles the geometry (SVG pattern manipulation, grading, SMPL body generation).
• Next.js + React + Three.js is the frontend.
• GSAP animates the cascade.

The brain-hands split: Claude reasons about what should happen, deterministic code executes how.

Development

See CLAUDE.md for the development workflow. In short:

1. Features start as specs in docs/specs/
2. /implement runs a TDD subagent that writes failing tests, then implementation
3. /review runs a fresh-context reviewer
4. All prompts are versioned files in prompts/
5. Prompts have evals in evals/

License

MIT — see LICENSE.

Credits

Built by Steph (Digital Smiles) for Anthropic's hackathon.