/library

The library.

Done-For-You CLAUDE.md files. Real templates you drop into your project root, edit a few lines, and ship. Every one is opinionated, voice-tested, and ready for Claude Code.

Web app

7 templates
  1. Next.js 16 SaaS starter (Stripe + magic-link auth)

    Opinionated CLAUDE.md for a Next.js 16 + Tailwind v4 SaaS with Stripe billing, Resend magic-link auth, and Vercel Postgres. Drop in your project root, edit five lines, ship.

    use when Starting a new SaaS with Next.js 16. You want billing + auth working day one without picking 12 libraries.

    2,680 bytesnextjstailwindstripesaasvercel
  2. Phoenix LiveView app (Elixir realtime, no JS framework)

    A full-stack Elixir app with Phoenix LiveView. Server-rendered, realtime updates over WebSocket, no React. Use when you want realtime collaboration features without the JS-state-management circus.

    use when Realtime dashboards, multiplayer apps, internal tools, anything where 'live updates from the server' is the core UX and you'd rather not maintain a separate frontend.

    2,240 byteselixirphoenixliveviewrealtimebeam
  3. Vite + React SPA with TanStack Query + Router

    A pure client-side React SPA on Vite. TanStack Query for server state, TanStack Router for type-safe routing, Zustand for local UI state. Use when SSR is overkill and you want a small fast bundle.

    use when Internal tool, admin dashboard, or product surface where SEO doesn't matter and you want React's full client-side power without Next.

    2,380 bytesvitereacttanstackspadashboard
  4. Go + Templ + HTMX (no-JS web app)

    A web app in pure Go. Templ for typed templates, HTMX for interactivity, SQLite for state. Single static binary deploys anywhere. Use when you want simplicity, speed, and a stack that won't have a churn cycle every six months.

    use when Internal tool, content site, admin dashboard, or product where the team has a backend-heavy skillset and 'a build step' is a red flag.

    2,210 bytesgohtmxtemplsqliteno-js
  5. Rails 8 + Hotwire + Tailwind (the boring stack that still ships)

    A Rails 8 app with Hotwire (Turbo + Stimulus), Tailwind v4, and Solid Queue / Solid Cache. SQLite in dev and prod-eligible thanks to Rails 8. Use when you want to ship a SaaS in a week with one process and zero JS framework debt.

    use when Solo-dev SaaS, internal tool, or any product where shipping in a week matters more than picking the trendy stack. Rails 8's omakase defaults eliminate 90% of the decisions.

    2,160 bytesrailsrubyhotwireturbostimulus
  6. Shopify Hydrogen storefront (custom headless e-commerce)

    A headless Shopify storefront on Hydrogen 2025 + Remix v3. Custom design, blazing-fast PDPs, Shop Pay still works. Use when Shopify themes can't express your brand and you want full design control without leaving the Shopify ecosystem.

    use when DTC brand outgrowing Shopify themes. You want custom design, control over performance, and your own deploy. Cart, checkout, and Shop Pay still go through Shopify.

    2,130 bytesshopifyhydrogenremixecommerceheadless

Tooling

6 templates
  1. Chrome MV3 extension with TypeScript + Vite

    A Manifest V3 extension built with Vite + TypeScript. Content script, background service worker, options page, and popup all bundled together. Cross-browser-ready (Chrome, Edge, Firefox).

    use when Browser extension that augments specific sites, captures data, or adds a side-panel tool. You want a real build system, not raw JS files.

    2,120 byteschromeextensionmv3vitebrowser
  2. Discord bot with discord.js + slash commands

    A TypeScript Discord bot with slash commands, button interactions, and modal forms. Deploy on Fly.io for $0-5 a month. Includes the awkward parts (command registration, gateway intents) so you don't have to figure them out.

    use when Community bot, moderation tool, embed-generator, or any in-server automation. You already use Discord and want one bot that does what you need.

    2,080 bytesdiscordbottypescriptfly
  3. Telegram bot (TypeScript + grammY framework)

    A TypeScript Telegram bot with grammY. Inline keyboards, webhook receive, conversation flows, payment buttons via Stars or Stripe. Use when your audience is on Telegram (crypto, gaming, EU regions).

    use when Channel automation, notification bot, Telegram-native product (sticker shop, mini-game, crypto-tipping). Discord doesn't reach this audience.

    2,010 bytestelegrambotgrammywebhook
  4. TypeScript CLI tool (zero-build, ESM, single file)

    A small TypeScript CLI you can run with `npx your-tool` or install globally. tsx for zero-build dev, commander for arg parsing, kleur for color, ora for spinners. Publishes to npm in two commands.

    use when Internal devtool, project scaffolder, one-off automation, or a public CLI you want to ship without fighting with bundlers.

    1,980 bytesclitypescriptnpmtsx
  5. Monorepo with pnpm workspaces + Turborepo (Next + Express + shared UI)

    A monorepo for teams running 2+ deployables (web app, marketing site, API, shared UI library, design tokens). pnpm workspaces, Turborepo for caching, changesets for versioning. Build the right packages, deploy the right apps, only when needed.

    use when Your team is shipping multiple apps sharing code (auth, UI, types). You're tired of copy-paste-modify between repos and ready to invest a day in monorepo infrastructure.

    2,090 bytesmonorepoturborepopnpmworkspaceschangesets
  6. Design system (shadcn/ui + Tailwind v4 + Storybook + tokens)

    A shippable design system. shadcn/ui as the base, Tailwind v4 tokens for theming, Storybook for documentation, Chromatic for visual regression. Use when your team needs a single source of truth for components across multiple products.

    use when You're a team of 3+ designers + devs. Components keep drifting between apps. You want a place where Button, Input, Dialog have one canonical version with one visual regression test suite.

    2,060 bytesdesign-systemshadcnstorybooktailwindchromatic

AI agents

4 templates
  1. RAG search service (Pinecone + OpenAI embeddings + Postgres)

    A document ingestion + retrieval-augmented generation service. Chunking, embeddings via OpenAI, vector storage in Pinecone, source metadata in Postgres, a typed query endpoint. Use when you need 'ChatGPT for your docs' but want to own the stack.

    use when You have a corpus of documents (support articles, internal wiki, product manuals, codebase) and you want users to ask questions and get answers grounded in the source material with citations.

    2,510 bytesragpineconeopenaiembeddingssearch
  2. Anthropic Computer Use agent (web automation in a sandbox)

    A Claude-driven agent that controls a sandboxed browser to complete real-world web tasks. Docker isolation, screenshot + click + type loop, task templates, run history. Use for QA bots, scraping, form filling, or anything 'go to a website and do a thing for me'.

    use when You need an agent that can actually USE websites the way a human would. Login, fill forms, click through flows, screenshot results. Not just parse HTML.

    2,360 bytesanthropiccomputer-useagentautomationsandbox
  3. Realtime voice agent (OpenAI Realtime API + Twilio)

    A phone-callable AI voice agent. Twilio handles the PSTN connection, OpenAI Realtime API handles speech-to-speech, your business logic intercepts via function calls. Use to replace 'press 1 for sales' phone trees with something that actually understands callers.

    use when Small business gets too many calls to staff a receptionist but every missed call is lost revenue. The agent handles common asks (hours, booking, quote requests) and warm-transfers the rest.

    2,440 bytesvoicetwilioopenairealtimephone
  4. LangGraph multi-agent orchestration (Python state-machine agents)

    A multi-step, multi-agent system built with LangGraph. Researcher + writer + critic agents collaborate via a typed state graph. Use when one giant prompt isn't cutting it and you need explicit handoffs, retries, and human-in-the-loop checkpoints.

    use when You're building a deep-research bot, an automated PR reviewer, a content pipeline, or any workflow where multiple AI calls + tool invocations need to coordinate with explicit state.

    2,330 byteslanggraphagentspythonorchestrationmulti-agent

Vertical (industry)

4 templates
  1. HVAC contractor website (lead form, service area, reviews, call tracking)

    A vertical-specific marketing site for an HVAC contractor. CallRail-style number, lead form to Resend/Twilio, service-area map with city pages, real-photo gallery, Google reviews embed. Built for the exact audience trades owners need to convert.

    use when You're building a website for an HVAC, plumbing, roofing, or similar trades client. The CLAUDE.md captures the specific decisions and gotchas of this vertical so Claude doesn't drift toward generic SaaS aesthetics.

    2,310 bytestradeshvaccontractorlead-formlocal-seo
  2. Photographer portfolio + client proofing galleries

    A portfolio site for a working photographer. Public galleries for marketing, password-gated client proofing galleries with download + favorite, inquiry form. Built for someone who shoots, not someone who codes.

    use when You're building for a wedding, portrait, or commercial photographer. Public galleries sell; private proofing galleries deliver. The CLAUDE.md captures both surfaces and the specific UX expectations.

    2,150 bytesphotographyportfoliogalleryclient-portal
  3. Restaurant menu + online ordering (Square integration)

    A site for a single restaurant. Live menu (synced from Square Catalog), online ordering for pickup, reservation link, locations + hours, real food photography. Built for restaurants that DON'T want to give 30% to DoorDash.

    use when You're building for a restaurant owner sick of 30% delivery fees. The site routes orders direct to Square; the kitchen prints them on the Square hardware they already own.

    2,120 bytesrestaurantsquareorderingmenuhospitality
  4. Real estate listings site (MLS feed + agent profiles + lead capture)

    A real estate brokerage site. Listings synced from an MLS feed (RETS/RESO Web API), agent profiles, lead form per listing, saved searches via magic-link auth. Built for a small-team brokerage that wants to escape Zillow rent.

    use when Independent brokerage or team that pays $1000+/mo to Zillow for leads and wants their own site that ranks for 'homes for sale in [city]' and routes leads to specific agents.

    2,230 bytesreal-estatemlsbrokeragelistingsleads

API / Backend

4 templates
  1. Node Express REST API with Drizzle + Zod

    A typed Node Express API. Drizzle ORM for Postgres, Zod schemas at the route boundary, JWT auth, Pino logging, Vitest for tests. Deploy anywhere a long-running Node process runs.

    use when Backend for a mobile app or SPA. You want a long-running Node process and a real ORM, not edge functions.

    2,360 bytesnodeexpressdrizzlezodpostgres
  2. Python FastAPI microservice with SQLAlchemy + Pydantic v2

    A small FastAPI service. Pydantic v2 schemas, SQLAlchemy 2.x async, Alembic migrations, structlog, pytest with httpx. Use when you want Python typing + speed without dragging in a framework.

    use when ML inference endpoint, data ingestion service, or any backend where Python's ecosystem (numpy, pandas, etc) is the reason you're not using Node.

    2,280 bytespythonfastapipydanticsqlalchemymicroservice
  3. Cloudflare Worker edge API with Hono + D1

    An edge API on Cloudflare Workers. Hono for routing, D1 for SQL at the edge, KV for sessions, Durable Objects when you need consistency. Sub-50ms responses globally, zero cold start.

    use when Auth proxy, webhook receiver, public API that needs to be fast everywhere, or anything where global latency matters more than database flexibility.

    2,270 bytescloudflareworkershonod1edge
  4. Stripe Checkout + Webhook receiver (drop-in payments)

    A minimal service that takes payments via Stripe Checkout and processes webhooks safely. Add to any project that needs one-time or subscription payments without building a full SaaS billing stack.

    use when You're selling something (a digital good, a license, a tier upgrade). You want Stripe Checkout for the UI, your webhook to update your DB, no third-party billing platform.

    2,180 bytesstripepaymentswebhookcheckout

Desktop

1 template
  1. Tauri 2 desktop app (Rust core + web frontend, native installer)

    A cross-platform desktop app with a Rust core and a Vite + React frontend. Smaller, faster, and more memory-efficient than Electron. Code-signed installers for macOS, Windows, Linux via GitHub Actions.

    use when Local-first tool, native menu bar app, file-system utility, or anything that needs to be a 'real' app on Mac/Windows/Linux without shipping a 200MB Electron binary.

    2,290 bytestaurirustdesktopvitereact

Content

1 template
  1. Astro 5 content site with MDX, RSS, and OG images

    A fast, content-led Astro 5 site. MDX with code highlighting, auto-generated OG images per post, RSS, sitemap, dark mode. Built for writing-heavy sites that need a 100 Lighthouse score.

    use when Personal site, technical blog, or company writing surface. You want shipping speed without giving up performance.

    2,540 bytesastromdxrsscontentblog

Mobile

1 template
  1. Expo React Native mobile app (iOS + Android + Web)

    A cross-platform Expo app with Expo Router (file-based), Tamagui for theming, EAS Build for releases, and Sentry for crash reporting. Single codebase ships to App Store, Play Store, and the web.

    use when Mobile-first product that also needs a web fallback. You don't want native Swift or Kotlin and you want OTA updates.

    2,470 bytesexporeact-nativetamaguieasmobile

Issues + case studies

1

New templates every week.

Subscribe and get each new CLAUDE.md in your inbox the day it ships.

also via RSS.