AI Engineering Stack Guide

This guide explains the modern AI engineering stack from the perspective of an AI solution architect and engineering lead: workflow frameworks, agent harnesses/runtimes, agent app frameworks, model serving, RAG/data, tools/MCP, evals, observability, security, and governance.

The site is intentionally ordered in this sequence:

Foundations: understand AI-DLC and Spec-Driven Development.
Decision tools: use the cheat sheet, wizard, templates, and scenario lab.
Stack map: understand model serving, RAG/data, tools/MCP, evals, security, and governance.
Agent layers: understand where LangChain, LangGraph, Hermes, Codex CLI, and Claude Code fit.
Deep dives: understand each framework on its own terms.
Comparison: compare only after the core context is clear.
Adoption: choose a stack for a real team, risk level, and codebase.

Start with the full stack

If you are confused because many frameworks look like plan -> implement -> review, start with the AI Engineering Stack Map. It explains which layer each framework owns before comparing individual tools.

One-Page Cheat Sheet

Fast layer map, selection matrix, and "do not compare directly" guide.

Decision Wizard

Interactive chooser that turns your context into a recommended stack.

Templates

Downloadable starter artifacts for specs, AI-DLC records, evals, and tool policies.

Scenario Lab

One RAG assistant feature implemented through different workflow lenses.

AI Engineering Stack

Full map of workflow, harness, app framework, model, RAG, tools, evals, and governance layers.

LangChain

Framework for building LLM apps and tool-calling agents.

LangGraph

Stateful orchestration framework for long-running agent systems.

GitHub Spec Kit

Spec-first delivery: intent becomes spec, plan, tasks, and implementation.

OpenSpec

Lightweight artifact-guided SDD with change folders, delta specs, and fluid iteration.

AWS AI-DLC Workflows

Lifecycle governance for AI-driven development with human approval and audit trail.

GSD / Get Shit Done

Context engineering and multi-agent execution for long-running delivery.

Superpowers

Engineering discipline skills: brainstorm, design, TDD, review, finish.

Hermes Agent

Open-source, hackable agent runtime/CLI for memory, tools, skills, and subagents.

Adjacent Ecosystem

Where OpenAI Agents SDK, AutoGen, CrewAI, Google ADK, Dify, n8n, and managed agents fit.

Quick orientation

If your biggest pain is...	Start with
You need the fastest decision path	One-Page Cheat Sheet
You want a stack recommendation from your context	Interactive Decision Wizard
You need copy-paste artifacts	Templates and Starter Artifacts
You want to see the same feature through each workflow	Scenario Lab
You are confused by adjacent agent tools	Adjacent Agent Ecosystem Map
You need the full AI engineering architecture map	Stack Map
You need production AI app quality	Evals & Observability
You need safe tool use and agent governance	Tools/MCP and Security/Governance
Requirements are vague and the agent guesses too much	GitHub Spec Kit
You want lightweight SDD without heavy gates	OpenSpec
Enterprise delivery needs approval, traceability, NFRs, and audit	AWS AI-DLC Workflows
The project spans many sessions and the agent loses context	GSD / Get Shit Done
The agent codes too quickly without tests, design, or review	Superpowers
You want to self-host or customize an agent runtime	Hermes Agent

mermaid

flowchart TB
    A[AI coding workflow landscape] --> B[Spec correctness]
    A --> AA[AI app orchestration]
    A --> C[Lifecycle governance]
    A --> D[Execution throughput]
    A --> E[Engineering discipline]
    A --> F[Agent runtime]

    B --> SK[GitHub Spec Kit]
    B --> OS[OpenSpec]
    C --> AD[AWS AI-DLC]
    D --> GSD[GSD]
    E --> SP[Superpowers]
    F --> HA[Hermes Agent]
    AA --> LC[LangChain]
    AA --> LG[LangGraph]

AI Engineering Stack Guide ​

Start with the full stack ​