AI harness for hardware. Describe the part you want — Fabware designs it, validates it against real manufacturing rules, exports a production-ready DXF for SendCutSend, and fills out the rest of your bill of materials from McMaster-Carr.
Chat left, design right. A "Lovable for hardware."
Early. Inherits a working prototype (chat → DSL → DXF → SCS handoff) and adds McMaster-Carr assembly parts. No auth yet — run locally.
See docs/PLAN.md for scope and roadmap.
lib/
api-spec/ OpenAPI source of truth (orval generates the rest)
api-zod/ Generated Zod schemas
api-client-react/ Generated React Query hooks
db/ Drizzle schema (Postgres)
integrations-anthropic-ai/
artifacts/
api-server/ Express API
hardwareai/ Vite + React + shadcn/ui workspace (chat + canvas)
mockup-sandbox/ Scratch
docs/
PLAN.md
Requires pnpm, Node 20+, and a Postgres database.
pnpm install
export DATABASE_URL=postgres://...
export ANTHROPIC_API_KEY=...
pnpm -r --filter @workspace/db run migrate # if migrations are set up
pnpm -r run buildDev:
pnpm --filter @workspace/api-server run dev
pnpm --filter @workspace/hardwareai run dev- SendCutSend — laser / waterjet / CNC flat-stock (DXF export, rule validation in
scsRules.ts) - McMaster-Carr — off-the-shelf fasteners, bearings, extrusion, etc. (part-number → deep-link, no scraping)
TBD.