shipanysaas is a Next.js and Supabase SaaS starter kit that Claude Code can work in without you explaining the codebase first. Its CLAUDE.md loads the rules in AGENTS.md, up to 49 ready-made skills sit in .claude/skills, and up to 17 written fixes cover the setup errors agents run into. Codex, Cursor and 7 more agents read the same rules.
Launch price: $49 web · $99 with iOS and Android, first 100 customers
Each item ships in the repository you get, and the docs explain how it works.
The files Claude Code reads before it writes a line, and the tests it runs after.
AGENTS.md lists what Claude Code runs after every change. The checks catch an agent’s mistakes before your users do.
Type check
The code has to compile with no type errors
Lint and format
Lint problems are fixed and the code is formatted
Adversarial review
The /reviewer skill tries to break the change
Data isolation test
After a database change, /rls-review tries to reach another account’s data
Pay once, accept the GitHub invitation, then build with your agent.
Starter is the web app; All-in adds iPhone, iPad and Android. At checkout you enter your GitHub username.
That account is invited to the private repository for your plan. Accept it, then clone the code to your computer.
Follow the setup guide once, then describe what you want. Claude Code reads the rules first and runs the checks before it says done.
Starter is the web app. All-in adds the iPhone, iPad and Android app on the same backend.
For founders and developers launching on the web.
Then $149 after the first 100 customers.
For founders and agencies who need iOS and Android too.
Then $249 after the first 100 customers.
Yes. Claude Code reads CLAUDE.md when a session starts, and the kit’s CLAUDE.md imports the root AGENTS.md. The skills sit in .claude/skills, where Claude Code finds them, and you can call one by name, such as /postgres-expert.
Yes. The repository is set up for Claude Code, GitHub Copilot, OpenAI Codex, Cursor, JetBrains Junie, Google Antigravity, Devin, Kiro, Grok Build, and Cline. Codex and Grok Build read AGENTS.md directly; the others have a short file that points to it. Every file is committed, so you can switch agents at any time.
AGENTS.md tells it to run the type check, lint and format, then the /reviewer skill, an adversarial review of the change. If the change touched the database, it also runs /rls-review, which writes tests that sign in as one account and try to reach another’s data.
AGENTS.md tells it to check the kit’s written fixes before guessing. Most have the error message, the cause and the change that worked. All-in has 17 fixes; Starter has the ones that are not about the phone app.
No, but you run a few commands once. The setup guide lists what to install (Node, pnpm, Docker and Git) and how to start the app. Then you describe what you want and Claude Code writes the code. Claude Code’s own plan is separate from the kit.
No. “For Claude Code” means the repository is set up for the agent you build with. The kit has no chat or AI features of its own. If your product needs them, your agent builds them on top.
Start tonight with the agent you already use. Launch, then submit it to the showcase.