Halden Solutions logoHalden Solutions wordmark

Agent Orchestration & Context

AGENTS.md Generator

Generate a universal AGENTS.md configuration file that provides unified repo context, guidelines, and execution boundaries to any AI coding agent.

Output: AGENTS.mdOpen Standard100% Client-Side

Start with a conversation

Talk to us

Use Halden AI as a sales assistant to explore services, understand offerings, and choose the right next step before booking a meeting.

Book a Meeting
Back to all toolsAGENTS.md Generator
Interactive Generator

Configure & Build your AGENTS.md

AGENTS.md is the universal open standard for giving autonomous coding agents the project context, technical stack details, build commands, and safety boundaries they need to build reliable code without hallucinations.

1. Project Identity

Set the project name and primary mission statement for autonomous coding agents.

2. Technology Stack Preset

Select your technical stack to automatically populate idiomatic CLI scripts, build tools, and framework rules.

3. Setup & Execution Commands

Explicitly tell AI agents which commands to invoke for dependency management, testing, and builds.

4. Engineering Standards & Safety Guardrails

Select the coding patterns, quality checks, and security rules that every agent PR must satisfy.

Code Style & Conventions

Testing Standards

Security & Secrets Protection

Add any proprietary conventions or architectural constraints.

AGENTS.md OutputSpec Ready

Ready to commit to your repository root.

# my-application

> Modern, production-grade enterprise application built with scalability and reliability in mind.

This file gives AI coding assistants (Claude Code, Cursor, Copilot, Codex, Aider, Devin) the technical context, architecture rules, and operational boundaries required to work effectively in this repository.

## Stack & Architecture
- **Framework**: Next.js (React · App Router · TypeScript)
- Use React Server Components by default; only add "use client" when state, effects, or browser APIs are required.
- Use Next.js Image component for images with proper priority and sizing.
- Group routes with parentheses e.g. (marketing) without affecting the URL structure.

## Commands & Scripts
- **Install dependencies**: `pnpm install`
- **Start development server**: `pnpm dev`
- **Build production bundle**: `pnpm build`
- **Run linter**: `pnpm lint`
- **Run automated tests**: `pnpm test`
- **Typecheck code**: `pnpm tsc --noEmit`

## Code Style & Conventions
- **TypeScript everywhere**: Use TypeScript for all new code. Avoid any — prefer unknown and narrow.
- **ESLint enforced**: Run lint before committing. Never disable rules without an explanatory comment.
- **Prettier formatting**: Auto-format on save. Do not format code by hand.
- **Tailwind CSS utility styling**: Use Tailwind classes. Extract components when UI patterns repeat 3+ times.
- **Strict naming conventions**: kebab-case files, PascalCase components, camelCase functions/variables, SCREAMING_SNAKE for env vars.
- **Sorted clean imports**: Group external → internal → relative. Use path aliases (@/*) instead of deep ../ relative imports.

## UI & Design System
- For all frontend styling, layout rules, and component creation, strictly follow the tokens and constraints defined in `./DESIGN.md`.
- Do not introduce arbitrary colors, fonts, or spacing outside the token scale in `./DESIGN.md`.

## Testing Standards
- **Unit testing with Vitest / Jest**: Co-locate *.test.ts files with source code. Test happy path and boundary edge cases.
- **E2E testing with Playwright**: Add end-to-end integration specifications for critical user journeys.
- **Accessibility (a11y) checks**: Enforce semantic HTML, ARIA attributes, keyboard navigation, and contrast validation.

## Security & Safety Boundaries
- **Never commit secrets or keys**: No API tokens or credentials in source code. Use .env.local (gitignored) and .env.example.
- **Schema validate all external input**: Validate all request bodies, params, and JSON payloads with schema validators (Zod / Valibot).
- **Server-side authorization**: Verify auth and permissions on the server for every endpoint. Do not rely solely on client guards.
- **Sanitize untrusted HTML**: Never render raw HTML from user input without sanitized escaping.

## Pull Request & Commit Rules
- **Conventional Commits**: Use feat:, fix:, chore:, refactor:, docs: with concise imperative summaries.
- **Small, focused changes**: One logical objective per pull request. Aim for clean, reviewable diffs.
- **Run checks before push**: Ensure lint, test, typecheck, and build pass cleanly before opening a PR.

## Custom Instructions
- Always explain architectural trade-offs when introducing new patterns.
- Prefer simplicity and maintainability over clever abstractions.

Integration

Where to place your AGENTS.md

Commit AGENTS.md to the root of your git repository. Most modern AI agents read this file on initial startup to understand the project structure and technical stack.

Repository File Structure
your-project/
├── AGENTS.md          ← root of repo (universal agent context)
├── DESIGN.md          ← optional UI tokens
├── package.json
└── src/

Key Capabilities & Features

  • Pre-configured smart presets for Next.js, Vite, Vue, SvelteKit, Astro, Node API, and Python
  • Configurable setup & script commands (install, dev, build, lint, test, typecheck)
  • Granular code style rules (TypeScript, ESLint, Prettier, Tailwind, component architecture)
  • Testing conventions and accessibility (a11y) guardrails
  • Security boundaries (secrets protection, input validation, server-side authentication)
  • Conventional commits & PR guidelines generator
  • One-click DESIGN.md integration toggle

Related Tags

Agent OrchestrationMulti-Agent SystemsCoding StandardsAI WorkflowsDeveloper Tooling

FAQ

Frequently Asked Questions

Everything you need to know about AGENTS.md and AI agent tooling.

What is AGENTS.md?

AGENTS.md is a standardized Markdown file placed in a repository root that provides AI agents with context about the tech stack, setup commands, coding standards, testing practices, and security rules.

How is AGENTS.md different from tool-specific rules like .cursor/rules or copilot-instructions?

AGENTS.md is a vendor-neutral, universal standard read by multiple AI coding tools (Claude Code, Cursor, Aider, Codex, Jules, Amp). You can maintain AGENTS.md as your single source of truth and link vendor-specific rules to it.

Can I include custom instructions for our proprietary architecture?

Yes. The generator provides customizable sections for directory structures, specialized API conventions, architectural patterns, and security constraints.

Explore

More Developer & AI Tools

Explore other open configuration generators from Halden Solutions.

DESIGN.mdDesign Systems & AI Rules

DESIGN.md Generator

Build a tokenized design system specification that AI coding agents can read, understand, and apply consistently across your codebase.

Launch Tool
mcp.jsonTool Calling & Agent Protocols

MCP Config Generator

Create Model Context Protocol (MCP) server configuration files to securely connect AI coding assistants to external tools, databases, APIs, and cloud resources.

Launch Tool
llms.txtAI SEO & Discoverability

llms.txt Generator

Create structured llms.txt and llms-full.txt files to guide AI search engines, LLMs, and agent web crawlers to your highest-value documentation.

Launch Tool
Enterprise Engineering

Need help building production AI platforms and multi-agent systems?

Halden Solutions helps teams design, build, and scale autonomous AI systems, secure agent guardrails, and enterprise cloud infrastructure.

Explore Cloud & Platform EngineeringGet in Touch