create-stitchkit

Create the canonical Stitchkit application:

bun create stitchkit my-app
cd my-app
# Point DATABASE_URL in .env at your PostgreSQL database.
bun run dev

The generated Bun workspace contains a Next.js frontend, a separate Stitchkit API, Prisma/PostgreSQL, typed shared contracts, Socket.IO, MCP, CLI tools and a complete production UI system.

To start from a terminal coding agent instead:

bun create stitchkit my-agent --template agent
cd my-agent
cp .env.example .env
# Set OPENROUTER_API_KEY, then choose a live model in the terminal.
bun run dev

This opt-in profile is deliberately smaller. It composes the published headless harness, Bun SQLite store, OpenRouter adapter, lazy filesystem skills and direct coding tools behind an OpenTUI shell. Reads/search run directly; writes, edits, patches and shell calls require a signed durable Y/N approval. bun --watch restarts source while the ignored .stitchkit/ directory retains conversation and recovery state. The bounded startup picker reads current tool-capable models and their context windows from OpenRouter instead of duplicating provider metadata in environment variables. The workspace path is a containment boundary, not an OS sandbox.

It uses one conventional packages/* namespace: backend, frontend, config, db and shared. The destination name becomes the generated slug; --display-name sets the human title. Both are recorded once in project.json — the generated project's declaration, the single machine-readable statement it makes about itself — and drive package, process, transport, UI and SEO identity.

The default scaffold is domain-free. To add the runnable repository example:

bun create stitchkit my-app --example repository

The optional example uses one configurable GitHub repository persisted through a server-side PostgreSQL cache, including a Prisma-backed visibility enum that flows through the shared schema and typed client. /en/ui presents every reusable primitive and composition. Every public page ships with localized metadata, canonical and language alternatives, sitemap coverage and a generated Open Graph card.

The included PM2 configs launch Bun and Next entrypoints directly, so package script wrappers do not hide the supervised processes or distort their metrics. PostgreSQL is supplied by the application environment through DATABASE_URL; the generated project does not package or start a database runtime.

The generated workspace uses the Stitchkit range declared once in the template catalog and ships with the exact lockfile validated by this package. Framework and scaffolder releases are independent: a new Stitchkit version does not move the starter until that target is deliberately updated and revalidated.