What is AAMAD ?

AAMAD is an open, production-grade framework for building multi‑agent AI applications that teaches teams how to define context, orchestrate specialized agents, and ship working applications with repeatable, auditable practices. It provides a complete virtual product development team of agent personas designed to run inside your Agentic Development Environment (i.e. Cursor or Claude Code) supplying the personas, prompts, rules, and artifacts that structure the work from idea to production.

AAMAD stands for AI‑Assisted Multi‑Agent Application Development and focuses on context engineering and spec-driven agentic development: it systematizes research, scoping, development, testing and delivery so teams can move from idea to production with traceability and repeatability. It emphasizes production‑ready architecture and documentation over demos, ensuring rapid prototyping without sacrificing governance, extensibility or quality.

Core ideas

• Persona‑driven development: Specialized development agent personas enforce single‑responsibility workflows in Cursor or similar IDEs, creating clear handoffs and auditable outputs across epics.
• Context as code: Every decision and step becomes a markdown artifact, making multi‑agent work explainable, reviewable, and reusable across projects and cohorts.
• Human backward‑compatibility: By requiring agents to document plans, status, and progress at each step, projects can be handed off or co‑developed with human engineers seamlessly.
• Development epics: Work is broken into epics that run in sequence or parallel, balancing speed with control while preserving artifact‑level traceability throughout.

Repository structure

• .cursor folder: for reusable framework assets (personas, prompts, rules, templates),
• project-context folder: for project‑specific outputs separated in Define, Build, and Deliver subfolders

This makes it easy to keep method and instance cleanly separated and allow agents to document epics without mixing framework assets and app artifacts.

Three phases at a glance

• Define: Produce Market Research and a PRD using templates so development agents later operate with shared context and explicit requirements.
• Build: Execute epics with AAMAD’s development agent personas to carry out architecture, frontend, backend, integration, and QA work.
• Deliver: Use the DevOps agent persona to build CI/CD pipelines and a containerized production environment to deploy your application.

Define Stage

Create a Market Research Document (MRD) and Product Requirements Document (PRD) with AAMAD templates to lock the problem frame and success criteria before coding begins.

Two common ways to do this:

• Option 1: Use the templates and prompts directly in a preferred AI tool (ChatGPT, Gemini, Claude, etc.) to generate MRD/PRD artifacts.
• Option 2: Install a Perplexity MCP server in Cursor, invoke the @product.mgr persona in a new chat, and request use of the Perplexity web search tool to drive research (requires Perplexity API).

Build Stage

AAMAD standardizes six epics, each documented in a specific markdown file to ensure consistent, reviewable outputs from development agent personas.

• Solution Architecture: sad.md for the target system design and decisions.
• Setup: setup.md for scaffolding, dependencies, and environment configuration.
• Frontend: frontend.md for UI and interaction surfaces and constraints.
• Backend: backend.md for services, data, and business logic.
• Integration: integration.md for wiring agents, tools, and services end‑to‑end.
• QA: qa.md for test plans, results, and limitations with follow‑ups.

AAMAD vs. CrewAI

A simple mental model is to separate a Development Crew from an Application Crew to avoid confusion at design time and runtime.

• AAMAD = Development Crew: Provides a full roster of development agent personas—Product Manager, Project Manager, System Architect, Frontend Engineer, Backend Engineer, Integration Engineer, QA Engineer, and DevOps Engineer—ready to be used in Cursor for orchestrated build cycles and artifact creation across phases.
• CrewAI = Application Crew: Recommended framework to implement and run the application agent personas that deliver value to end users (tools, tasks, and orchestrations specific to the product’s job‑to‑be‑done).

Your Role

Throughout the course, you will take the role of the Agentic Architect who coordinates the Development Crew (AAMAD) and designs the Application Crew (CrewAI) to deliver real customer value with clear specs, quality gates, and operational readiness baked in.

How to get started

• Clone the repo into your project directory [GitHub Repo]

• Make sure the .cursor rules are loaded in the agent chat window

• Get to know your product development team!

  • invoke one of the agent personas and ask to explain what it can do ("@project.mgr tell me what you can do"), then explore the epics and artifacts it drives.