AI Prompt Builder
Structure prompts with a role, task, context, constraints, output format and examples, with reusable variables.
Tell AI coding assistants how your project works: the stack, the exact commands, where things live and the conventions you care about. Write it once, then export the file each tool reads.
How it works
Pick a stack preset or start from scratch, and describe the project in a sentence or two.
Add the exact install, dev, build, test and lint commands, and the main folders.
Tick the conventions you want and add your own rules and don'ts.
Choose a tool to preview its file, then copy it, download it, or download all files as a zip.
About this tool
Coding agents start every session knowing nothing about your project. A short rules file at the root of the repository fixes that: which package manager and test command to use, where routes and shared code live, which patterns to follow and which files never to touch. Agents read it automatically, so you stop repeating the same instructions and get changes that look like the rest of your code.
Each tool looks for its own file. AGENTS.md is an open format read by a growing list of agents; Claude Code reads CLAUDE.md; Cursor uses rule files in .cursor/rules; GitHub Copilot reads .github/copilot-instructions.md; Windsurf and Gemini CLI have their own. This generator writes the same content for each, with the front matter each one expects. Keep it short and specific: rules are added to every request, so a page of essentials works better than a style guide.
AGENTS.md, CLAUDE.md, Cursor .mdc, Copilot instructions, Windsurf rules and GEMINI.md from one form.
Start from Next.js, Astro, FastAPI, Django, Rails or Go with sensible commands and structure.
Common rules such as focused diffs, strict types and tests with every change, written clearly.
Get every file in a zip, already in the right folders.
FAQ
Use the one your tool reads. If your team uses several tools, keep AGENTS.md as the main file and, for Claude Code, use the option that makes CLAUDE.md import it with @AGENTS.md so there's one source of truth.
What an agent can't work out from the code: exact commands, project layout, conventions that differ from defaults, and things to avoid. Skip general advice like "write clean code".
Short. The file is included with every request, so long files cost tokens and dilute the important rules. A page or two is plenty; the generator shows the word and token count.
Yes. Most tools also read nested files (an AGENTS.md or CLAUDE.md in a subfolder), and Cursor rules can target file patterns with globs. Start with one project-wide file.
Missing something?
Describe the job and the file types involved. The most requested tools get built first.