CLAUDE.md generator

Pick your stack, package manager, and team rules: get a copyable CLAUDE.md built for your project.

Step 1 / 4

What's the project's stack?

Want this inside your own product?

Let's talk in a 45-min discovery call.

Book a call

Most CLAUDE.md examples online share the same shape: project name, a few generic rules, some style preferences. The two sections that actually change an agent's behavior are usually missing entirely: a mandatory validation rule that names the command a task must pass before it counts as done, and an explicit list of what never happens without asking. Skip both and the agent decides on its own.

This tool ships real default commands for four stacks. Pick Next.js plus npm and the validation command fills in as npm run build && npx tsc --noEmit. Pick Python and you get pytest -q for validation and ruff check . for lint. Switch package managers and the commands (npm run build, pnpm build, yarn build, bun run build) update automatically, none of it is a blank template. Our Cursor, Claude Code, and Windsurf comparison covers more on running a team on Claude Code.

A CLAUDE.md file governs how an agent behaves, it doesn't guarantee the code itself. Need an RLS policy written? See our Supabase RLS policy generator. If an agent-built codebase grows past what the file can cover once real users show up, our AI project completion service picks up from there.

Frequently asked

Does this tool read my files or connect to my project?

No. It doesn't read any file or connect to a server. It generates the CLAUDE.md text in your browser from the stack, package manager, and rules you pick, and none of that data goes to Lumiasoft or anywhere else.

Why can I edit the commands?

Every stack ships a real default, for example picking Next.js plus npm fills the validation command in as npm run build && npx tsc --noEmit. But if your project's actual commands differ, you can edit the field and type your own.

Why is the validation command a mandatory section?

Because it's the part of a CLAUDE.md file that does the most work: if you don't name the command that has to pass before a task counts as done, the agent decides on its own whether to skip it. This tool puts that rule at the center of the file.

Is the generated file specific to my project, or just a template?

Both: the section headings and the validation rule stay fixed, but the stack, package manager, commands, language, and the operations that require approval all change based on your picks. Any command field you leave blank fills in with the right default for your stack.