AGENTS.md generator for AI coding agents

Generate an AGENTS.md for your next project

Define your stack, architecture, coding style, and agent instructions in minutes. Build a clear project setup for AI coding agents like Codex, Claude, and Cursor without a login or API key.

AGENTS.md

# AGENTS.md

## Project Overview
AgentSeed is a frontend SPA for generating project instructions.

## Agent Workflow
- Inspect relevant files before editing.
- Make minimal, targeted changes.
- Explain validation before final response.

## Do Not Do Without Approval
- Do not introduce major dependencies.
- Do not change architecture broadly.

Features

A focused tool for project instructions

AgentSeed helps developers create reliable project context for AI coding agents without adding process overhead.

Agent-ready project context

Turn your stack, architecture, testing rules, and project instructions into a clear AGENTS.md baseline.

Opinionated but editable

Start with sensible defaults, customize every important rule, then refine the markdown in your repository.

Copy or download instantly

Use the generated file immediately with one-click copy or a valid AGENTS.md download.

No login, no setup

Everything runs locally in the browser with deterministic generation and no external AI API.

Generator

Generate project instructions live

Configure the project type, stack, coding standards, testing expectations, and approval rules. The AGENTS.md preview updates immediately.

Configure your project

Start with the defaults, then tune the project instructions to match your repository.

Project details

Primary stack

Architecture

Code style preferences

Testing expectations

Agent behavior rules

Initial setup prompts

Add one-time setup tasks agents should complete when initializing the project, then remove after completion.

Extra agent habits

Add optional final-response habits that make agent work easier to review.

Output preferences

Include sections

Live AGENTS.md Preview

4,349 characters generated

# AGENTS.md

## Project Overview
This project is a Frontend SPA. A software project maintained with help from AI coding agents.

Use this file as the operating guide for AI coding agents working in this repository.

## Tech Stack
- Nuxt / Vue
- TypeScript
- Tailwind CSS

## Architecture
The preferred architecture style is Feature-based.

- Respect the existing module boundaries and folder conventions before adding new structure.
- Keep related code close to the feature or layer it belongs to.
- Avoid introducing new architectural patterns unless the change clearly requires it and the user approves.
- When uncertain, inspect similar existing implementations and follow the local precedent.

## Coding Standards
Keep communication direct and focused on the change, validation, and any important tradeoffs.

- Prefer strict TypeScript and avoid weakening types for convenience.
- Keep components small, focused, and easy to scan.
- Use meaningful names for variables, functions, components, files, and database objects.
- Avoid premature abstraction; duplicate small pieces until a pattern is clear.
- Follow existing patterns before introducing new libraries, folders, or conventions.

## Agent Workflow
- Inspect the relevant files and surrounding patterns before making changes.
- Make minimal, targeted changes that directly address the request.
- Do not rewrite, rename, or reformat unrelated code.
- Ask for explicit approval before broad architecture or data model changes.
- Update AGENTS.md when project conventions, architecture, or agent instructions change.
- Never commit secrets, tokens, credentials, or private environment values.
- Preserve existing behavior unless asked otherwise.
- Follow established naming, formatting, and folder conventions.
- Explain tradeoffs when making non-obvious decisions.

## Testing & Validation
- Add or update unit tests for business logic and pure functions.
- Run the relevant test, lint, typecheck, or build commands before the final response.
- If validation cannot be run, explain why and describe the remaining risk.

## Documentation Rules
- Update README.md when setup, commands, environment variables, or usage changes.
- Update AGENTS.md when project conventions, architecture, or agent instructions change.
- Document new public APIs, components, or non-obvious behavior.

## Extra Agent Habits
- After making changes, suggest a conventional commit message that accurately summarizes the work.
- Mention the commands run for validation and summarize the result.
- Call out meaningful risks, assumptions, or areas that need user review.

## Initial Setup Prompts
These are one-time setup prompts for agents initializing the project. Complete relevant tasks early, then remove completed items from this section so the file stays current.

- Polish the project description in this file. If the current description is vague, ask clarifying questions before rewriting it.
- Create a `CLAUDE.md` file whose content is exactly `@AGENTS.md` so Claude-based agents reuse these instructions.
- Confirm the setup, development, test, and build commands from package/config files and update documentation if they differ from this file.

## Security Rules
- Never commit secrets.
- Use environment variables for credentials.
- Validate input at boundaries.
- Avoid logging sensitive data.

## UI / UX Rules
- Keep UI accessible.
- Use semantic HTML.
- Preserve responsive behavior.
- Avoid unnecessary visual complexity.

## SEO Rules
- Keep page titles and meta descriptions meaningful.
- Use semantic headings.
- Add Open Graph and Twitter metadata where appropriate.
- Avoid client-only rendering for SEO-critical content when possible.

## Performance Rules
- Avoid unnecessary dependencies.
- Lazy-load heavy UI where useful.
- Keep bundle size in mind.
- Prefer simple solutions first.

## Git & Pull Request Guidelines
- Keep changes scoped.
- Summarize what changed.
- Mention tests run.
- Call out follow-up work separately.

## Do Not Do Without Approval
- Do not introduce major dependencies.
- Do not change architecture broadly.
- Do not remove tests.
- Do not reformat unrelated files.
- Do not alter CI/CD or deployment settings unless asked.

## Maintenance
- Revisit this file when conventions change.
- Keep instructions specific to the actual codebase.
- Remove outdated rules.

FAQ

Common questions

What is AGENTS.md?

AGENTS.md is a project instructions file for AI coding agents. It explains your stack, conventions, workflow rules, testing expectations, and approval boundaries.

Which agents does this work with?

AgentSeed is useful for AI coding agents that can read repository instructions, including Codex, Claude, Cursor, and similar tools.

Should I commit AGENTS.md to my repo?

Usually, yes. Committing it keeps project instructions visible, reviewable, and consistent for everyone using the codebase.

Can I edit the generated file?

Yes. The generated markdown is meant to be copied, downloaded, committed, and refined as your project conventions evolve.

Does it use an AI API?

No. The MVP uses deterministic client-side generation, so there is no external AI API key, account, or project upload required.

Can I suggest an improvement?

Yes. If you have an idea to improve the generator, send it to agentseed@dion.software.