Prompthon-IO/agent-systems-handbook

Agent Systems Handbook by Prompthon — A practical AI agents handbook covering agent systems, agentic workflows, LangGraph, MCP/A2A, context engineering, agent memory, evaluation, observability, and multi-agent architecture. Current trend focus: Gemini Interactions API and managed agents, emerging agent runtimes, and production AI workflow patterns.

What it solves

AI-agent demos are easy to find, but production-ready agent systems are harder to understand. This handbook closes that gap by mapping the workflows, tools, memory systems, context engineering, interoperability (MCP/A2A), evaluation, observability, and multi-agent architecture behind real-world AI agents. It gives readers a structured way to learn, design, build, and operate such systems from first principles.

How it works

The handbook is organized into four parallel reading paths — Explorer, Practitioner, Builder, and Contributor — so different types of learners can follow the track that fits their goals rather than a single sequence. Content spans AI agent foundations, agentic workflows, planning, reflection, tool use, agent memory and retrieval, context engineering, agentic RAG, protocols like MCP and A2A, frameworks such as LangGraph, multi-agent orchestration, evaluation, observability, reliability, and safety. It also includes deep research agents, customer-support agents, source projects, and starter examples. The material is created with an AI-native workflow (AI-assisted drafting, synthesis, and refinement) combined with expert review.

Who it’s for

  • Students, newcomers, and curious readers who want a broad understanding of AI agents and trends (Explorer)
  • People who want to use AI tools and workflows in daily life, study, or work without becoming full-time engineers (Practitioner)
  • Engineering-minded learners, new grads, and developers who want to build agent applications and systems (Builder)
  • Those who want to contribute to the lab by adding, revising, or maintaining content (Contributor)

Highlights

  • Four parallel reading paths (Explorer, Practitioner, Builder, Contributor) so each reader can choose their own track
  • Covers agent foundations, agentic workflows, planning, reflection, tool use, memory, RAG, context engineering, MCP/A2A, multi-agent orchestration, evaluation, observability, and safety
  • AI-native content created through an assisted workflow with expert guidance
  • Public contribution paths with templates and review standards for lab articles, radar notes, source projects, skills, and reference notes

Related

  • Project
    DAWNCR0W/affine-mcp-server

    An MCP server that enables AI assistants to interact with AFFiNE workspaces, allowing for automated document creation, database management, and workspace organization.

  • Project
    workopia/workopia-mcp

    A hosted MCP server that integrates job searching, resume tailoring, and application tracking into AI assistants like Claude and Cursor.

  • Project
    jd-opensource/JoySafeter

    An AI-native platform for building and orchestrating security agents to automate complex tasks like vulnerability analysis and penetration testing.

  • Project
    mini-yifan/deepseek-harness-orb

    DeepSeek Orb is a desktop AI agent for macOS and Windows that uses Computer Use technology to perform visible GUI actions and manage complex background tasks from a persistent, screen-edge interface.