create-ai-agents-docs

Scaffold AI agent-friendly documentation into any repository with a single command.

Usage

npx create-ai-agents-docs

This runs in interactive mode and guides you through template selection and output directory.

Non-interactive

# Use the built-in default template
npx create-ai-agents-docs --template default

# Use a template from a GitHub repository
npx create-ai-agents-docs --template https://github.com/some-org/some-repo

# Use a template from a local path
npx create-ai-agents-docs --template ./my-custom-template

# Specify the output directory (defaults to current directory)
npx create-ai-agents-docs --output ./my-project

Flags

Flag	Description	Default
--template	Template name, GitHub URL, or local path	default
--output	Directory to scaffold files into	current dir
--force	Overwrite existing files	false
--help	Show usage information
--version	Show version number

Built-in templates

default

A comprehensive starter for AI agent-friendly documentation. Includes:

• AGENTS.md — instructions and context for AI agents working in the repository
• CLAUDE.md — symlink to AGENTS.md (Claude Code)
• .cursorrules — symlink to AGENTS.md (Cursor)
• .windsurfrules — symlink to AGENTS.md (Windsurf)
• .clinerules — symlink to AGENTS.md (Cline)
• GEMINI.md — symlink to AGENTS.md (Gemini CLI)
• .github/copilot-instructions.md — symlink to AGENTS.md (GitHub Copilot)
• CONTRIBUTING.md — contribution guidelines
• docs/GLOSSARY.md — project terminology
• docs/architecture/README.md — architecture overview
• docs/adr/adr-template.md — Architecture Decision Record template
• .claude/settings.json — Claude Code configuration

microservices

For distributed systems composed of multiple independently deployable services. Prompts for project name and communication pattern (REST/gRPC/events/mixed). Includes:

• AGENTS.md — service map, data ownership rules, inter-service communication patterns, cross-service change rollout process, observability requirements
• CONTRIBUTING.md — microservices contribution workflow including contract change coordination
• docs/contracts/README.md — API and event contract documentation templates with versioning guidance
• docs/GLOSSARY.md — distributed systems terminology (bounded context, correlation ID, idempotent, saga, service contract)
• docs/architecture/README.md — service map, data ownership, communication patterns, observability stack
• docs/adr/adr-template.md — Architecture Decision Record template
• .claude/settings.json — Claude Code configuration
• Agent symlinks (CLAUDE.md, .cursorrules, .windsurfrules, .clinerules, GEMINI.md, .github/copilot-instructions.md)

cli

For command-line tools and developer utilities written in TypeScript, Go, Rust, or Python. Prompts for project name and language. Includes:

• AGENTS.md — command/subcommand structure, stdout vs. stderr conventions, exit code rules, configuration resolution order, how to add a new command
• CONTRIBUTING.md — CLI contribution workflow including output behavior verification
• docs/commands/README.md — command reference template with flags, arguments, and exit codes
• docs/GLOSSARY.md — CLI terminology (command, subcommand, flag, argument, stdout, stderr, exit code)
• docs/architecture/README.md — command tree, argument parsing, configuration resolution, plugin system
• docs/adr/adr-template.md — Architecture Decision Record template
• .claude/settings.json — Claude Code configuration
• Agent symlinks (CLAUDE.md, .cursorrules, .windsurfrules, .clinerules, GEMINI.md, .github/copilot-instructions.md)

fullstack

For full-stack web applications with a frontend, backend, and database in one repository. Prompts for project name, frontend framework, and backend language. Includes:

• AGENTS.md — client/server boundary rules, shared code conventions, API contract discipline, environment variable safety, database migration workflow
• CONTRIBUTING.md — full-stack contribution workflow including cross-layer change guidance
• docs/GLOSSARY.md — full-stack terminology (client, server, shared, API contract)
• docs/architecture/README.md — system diagram, auth flow, data fetching strategy, state management, database
• docs/adr/adr-template.md — Architecture Decision Record template
• .claude/settings.json — Claude Code configuration
• Agent symlinks (CLAUDE.md, .cursorrules, .windsurfrules, .clinerules, GEMINI.md, .github/copilot-instructions.md)

design-system

For component libraries and design systems built with React, Vue, Web Components, or Svelte. Prompts for project name and framework. Includes:

• AGENTS.md — component authoring guidelines, WCAG 2.1 AA accessibility requirements, design token rules, Storybook story requirements, visual regression testing
• CONTRIBUTING.md — design system contribution workflow including an accessibility checklist
• docs/components/README.md — component inventory with status tracking (Stable/Beta/Deprecated)
• docs/GLOSSARY.md — design system terminology (tokens, variants, slots, visual regression)
• docs/architecture/README.md — token hierarchy, component composition model, theming, component categories
• docs/adr/adr-template.md — Architecture Decision Record template
• .claude/settings.json — Claude Code configuration
• Agent symlinks (CLAUDE.md, .cursorrules, .windsurfrules, .clinerules, GEMINI.md, .github/copilot-instructions.md)

data-pipeline

For data engineering projects — ETL/ELT pipelines, data warehouses, dbt, Airflow, Dagster, or Spark. Prompts for project name, orchestrator, and transformation tool. Includes:

• AGENTS.md — idempotency rules, backfill guidance, data quality requirements, schema change process, pipeline conventions
• CONTRIBUTING.md — data pipeline contribution workflow including backfill and schema change documentation
• docs/GLOSSARY.md — data engineering terminology (DAG, backfill, idempotent, raw/staging/mart layers)
• docs/architecture/README.md — data flow diagram, source systems, pipeline inventory, data quality strategy
• docs/adr/adr-template.md — Architecture Decision Record template
• .claude/settings.json — Claude Code configuration
• Agent symlinks (CLAUDE.md, .cursorrules, .windsurfrules, .clinerules, GEMINI.md, .github/copilot-instructions.md)

infrastructure

For Infrastructure-as-Code projects — Terraform, Pulumi, CloudFormation, or Kubernetes. Prompts for project name and IaC tool. Includes:

• AGENTS.md — plan-before-apply workflow, environment hierarchy (dev/staging/prod), state management rules, naming conventions, blast radius awareness
• CONTRIBUTING.md — IaC contribution workflow including staged rollout and peer review requirements
• docs/GLOSSARY.md — infrastructure terminology (blast radius, drift, idempotent, plan, state)
• docs/architecture/README.md — environment topology, network layout, compute resources, IAM, secrets management
• docs/adr/adr-template.md — Architecture Decision Record template
• .claude/settings.json — Claude Code configuration
• Agent symlinks (CLAUDE.md, .cursorrules, .windsurfrules, .clinerules, GEMINI.md, .github/copilot-instructions.md)

mobile

For mobile applications built with React Native, Flutter, SwiftUI, or Kotlin. Prompts for project name and framework. Includes:

• AGENTS.md — platform-specific guidance (iOS/Android), build commands, simulator/emulator setup, navigation structure, platform-conditional code patterns
• CONTRIBUTING.md — mobile contribution workflow including cross-platform testing requirements
• docs/GLOSSARY.md — mobile terminology (bridge, native module, simulator, emulator)
• docs/architecture/README.md — navigation structure, state management, data fetching, native modules
• docs/adr/adr-template.md — Architecture Decision Record template
• .claude/settings.json — Claude Code configuration
• Agent symlinks (CLAUDE.md, .cursorrules, .windsurfrules, .clinerules, GEMINI.md, .github/copilot-instructions.md)

library

For reusable packages published to a registry (npm, PyPI, crates.io, Maven). Prompts for package name and registry. Includes:

• AGENTS.md — public API rules, semantic versioning policy, deprecation process, how to add a new feature without breaking changes
• CONTRIBUTING.md — library contribution workflow including breaking change guidance
• docs/api/README.md — public API reference template
• docs/GLOSSARY.md — library terminology (public API, consumer, breaking change)
• docs/architecture/README.md — module structure, public vs. internal, extension points
• docs/adr/adr-template.md — Architecture Decision Record template
• .claude/settings.json — Claude Code configuration
• Agent symlinks (CLAUDE.md, .cursorrules, .windsurfrules, .clinerules, GEMINI.md, .github/copilot-instructions.md)

api-service

For backend API services — REST or GraphQL, with a database, auth, and deployment. Prompts for project name and API style. Includes:

• AGENTS.md — API conventions, request validation rules, auth patterns, database migration workflow, how to add a new endpoint
• CONTRIBUTING.md — API-focused contribution workflow including migration guidance
• docs/api/README.md — endpoint documentation template
• docs/GLOSSARY.md — project terminology
• docs/architecture/README.md — request lifecycle, service layers, data model, auth
• docs/adr/adr-template.md — Architecture Decision Record template
• .claude/settings.json — Claude Code configuration
• Agent symlinks (CLAUDE.md, .cursorrules, .windsurfrules, .clinerules, GEMINI.md, .github/copilot-instructions.md)

python

For Python projects — web apps (Django, Flask, FastAPI), scripts, data science, or libraries. Prompts for project name and package manager (pip/poetry/uv). Includes:

• AGENTS.md — virtual environment setup, pytest usage, ruff/mypy workflow, type annotation and import style rules
• CONTRIBUTING.md — Python-specific contribution workflow including the full check suite
• docs/GLOSSARY.md — project terminology
• docs/architecture/README.md — architecture overview
• docs/adr/adr-template.md — Architecture Decision Record template
• .claude/settings.json — Claude Code configuration
• Agent symlinks (CLAUDE.md, .cursorrules, .windsurfrules, .clinerules, GEMINI.md, .github/copilot-instructions.md)

monorepo

For multi-package workspaces managed with Nx, Turborepo, pnpm workspaces, npm workspaces, or Yarn workspaces. Prompts for project name and workspace tool. Includes:

• AGENTS.md — workspace map, package boundary rules, how to run commands at workspace root vs. package level, and how to add new packages
• CONTRIBUTING.md — monorepo contribution workflow, cross-package change guidance
• docs/GLOSSARY.md — workspace terminology (app, package, root, workspace)
• docs/architecture/README.md — package dependency graph, shared packages, deployment units
• docs/adr/adr-template.md — Architecture Decision Record template
• .claude/settings.json — Claude Code configuration
• Agent symlinks (CLAUDE.md, .cursorrules, .windsurfrules, .clinerules, GEMINI.md, .github/copilot-instructions.md)

External templates

Point --template at any GitHub repository or local directory that follows the template structure. GitHub repositories are cloned with git clone --depth 1; local paths are used directly.

npx create-ai-agents-docs --template https://github.com/org/my-template
npx create-ai-agents-docs --template ./path/to/my-template
npx create-ai-agents-docs --template /absolute/path/to/my-template

Template authoring

To build a custom template, use the create-ai-agents-docs-template starter kit — it contains the default template's files as a starting point.

A template is a directory of files to scaffold. Any files in the directory are copied to the target when the template is applied.

Template structure

my-template/
├── template.json   # optional config (not copied to target)
├── AGENTS.md
├── CONTRIBUTING.md
└── docs/
    └── GLOSSARY.md

template.json

Templates may include a template.json at their root to configure variables, post-copy hooks, and ignored files. All fields are optional.

{
  "name": "My Custom Template",
  "description": "A short description shown in interactive mode",
  "variables": {
    "projectName": {
      "prompt": "What is your project name?",
      "default": "my-project"
    },
    "author": {
      "prompt": "Author name?"
    }
  },
  "hooks": {
    "postCopy": "echo 'Done! Run npm install to get started.'"
  },
  "ignore": ["README-template.md"]
}

Variables

Variables prompt the user for values and substitute {{variableName}} placeholders in file contents.

# {{projectName}}

Maintained by {{author}}.

If a variable has a default, it is used when the user presses Enter without typing a value.

Post-copy hooks

hooks.postCopy is a shell command (or semicolon-separated commands) run in the target directory after files are copied.

Ignore list

ignore is a list of file paths relative to the template root that should not be copied to the target. template.json is always excluded automatically.

Tips

• Keep templates focused. A template should scaffold a specific kind of project, not try to cover every use case.
• Use {{variableName}} placeholders for values that differ between projects (project name, author, etc.).
• The description field in template.json is shown in interactive mode — keep it under 60 characters.
• Test your template locally before publishing: npx create-ai-agents-docs --template ./my-template --output /tmp/test-output

License

Licensed under MIT. Do what you will.