APM – Agent Package Manager

The dependency manager for AI agents. apm.yml declares the skills, prompts, instructions, and tools your project needs — so every developer gets the same agent setup. Packages can depend on packages, and APM resolves the full tree.

Think package.json, requirements.txt, or Cargo.toml — but for AI agent configuration.

GitHub Copilot · Cursor · Claude · Codex · Gemini

Why APM

AI coding agents need context to be useful: what standards to follow, what prompts to use, what skills to leverage. Today this is manual — each developer installs things one by one, writes instructions from scratch, copies files around. None of it is portable. There's no manifest for it.

APM fixes this. You declare your project's agentic dependencies once, and every developer who clones your repo gets a fully configured agent setup in seconds. Packages can depend on other packages — APM resolves transitive dependencies automatically, just like npm or pip.

See It in Action

# apm.yml — ships with your project, like package.json
name: corporate-website
dependencies:
  apm:
    - danielmeppiel/form-builder       # Skills: React Hook Form + Zod
    - danielmeppiel/compliance-rules   # Guardrails: GDPR, security audits
    - danielmeppiel/design-guidelines  # Standards: UI consistency, a11y

New developer joins the team:

git clone your-org/corporate-website
cd corporate-website
apm install && apm compile

That's it. Copilot, Claude, Cursor — every agent is configured with the right skills, prompts, and coding standards. No wiki. No "ask Sarah which skills to install." It just works.

→ View the full example project

Not Just Skills

Skill registries install skills. APM manages every primitive your AI agents need:

Primitive	What it does	Example
Instructions	Coding standards, guardrails	"Use type hints in all Python files"
Skills	AI capabilities, workflows	Form builder, code reviewer
Prompts	Reusable slash commands	/security-audit, /design-review
Agents	Specialized personas	Accessibility auditor, API designer
MCP Servers	Tool integrations	Database access, API connectors

All declared in one manifest. All installed with one command — including transitive dependencies:

apm install → integrates prompts, agents, and skills into .github/ and .claude/ apm compile → compiles instructions into AGENTS.md (Copilot, Cursor, Codex) and CLAUDE.md (Claude)

Get Started

1. Install APM

curl -sSL https://raw.githubusercontent.com/danielmeppiel/apm/main/install.sh | sh

Homebrew or pip

brew tap danielmeppiel/apm-cli && brew install apm-cli
# or
pip install apm-cli

2. Add packages to your project

apm install danielmeppiel/compliance-rules

No apm.yml yet? APM creates one automatically on first install.

3. Compile your instructions

apm compile

Done. Your instructions are compiled into AGENTS.md and CLAUDE.md — open your project in VS Code or Claude and your agents are ready.

Install From Anywhere

apm install owner/repo                                              # GitHub
apm install github/awesome-copilot/prompts/code-review.prompt.md   # Single file
apm install ghe.company.com/owner/repo                             # GitHub Enterprise
apm install dev.azure.com/org/project/repo                         # Azure DevOps

Create & Share Packages

apm init my-standards && cd my-standards

my-standards/
├── apm.yml              # Package manifest
└── .apm/
    ├── instructions/    # Guardrails (.instructions.md)
    ├── prompts/         # Slash commands (.prompt.md)
    └── agents/          # Personas (.agent.md)

Add a guardrail and publish:

cat > .apm/instructions/python.instructions.md << 'EOF'
---
applyTo: "**/*.py"
---
# Python Standards
- Use type hints for all functions
- Follow PEP 8 style guidelines
EOF

git add . && git commit -m "Initial standards" && git push

Anyone can now apm install you/my-standards.

All Commands

Command	What it does
apm install <pkg>	Add a package and integrate its primitives
apm compile	Compile instructions into AGENTS.md / CLAUDE.md
apm init [name]	Scaffold a new APM project or package
apm run <prompt>	Execute a prompt workflow via AI runtime
apm deps list	Show installed packages and versions
apm compile --target	Target a specific agent (vscode, claude, all)

Configuration

For private repos or Azure DevOps, set a token:

Token	When you need it
GITHUB_APM_PAT	Private GitHub packages
ADO_APM_PAT	Azure DevOps packages
GITHUB_COPILOT_PAT	Running prompts via apm run

→ Complete setup guide

---

Community Packages

APM installs from any GitHub or Azure DevOps repo — no special packaging required. Point at a prompt file, a skill, or a full package. These are some curated packages to get you started:

Package	What you get
danielmeppiel/compliance-rules	/gdpr-assessment, /security-audit + compliance guardrails
danielmeppiel/design-guidelines	/accessibility-audit, /design-review + UI standards
DevExpGbb/platform-mode	Platform engineering prompts & agents
github/awesome-copilot	Community prompts, agents & instructions for Copilot
anthropics/courses	Anthropic's official skills & prompt library
Add yours →

---

Documentation

Get Started	Quick Start · Core Concepts · Examples
Reference	CLI Reference · Compilation Engine · Skills · Integrations
Advanced	Dependencies · Primitives · Contributing

---

Built on open standards: AGENTS.md · Agent Skills · MCP

Learn AI-Native Development → Awesome AI Native A practical learning path for AI-Native Development, leveraging APM along the way.

Trademarks

This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.