notes
Agentic Design Patterns
Reference
Source: Agentic Design Patterns by Antonio Gulli (424 pages)
A hands-on study guide for building intelligent autonomous systems that perceive, decide, and act with minimal human intervention.
Core Patterns (Part One)
| Pattern | Description |
|---|---|
| Prompt Chaining | Decompose tasks into sequential LLM calls |
| Routing | Direct inputs to specialized handlers |
| Parallelization | Fan-out/fan-in workloads across agents |
| Reflection | Self-critique and revision loops |
| Tool Use | Agents calling external tools and APIs |
| Planning | Multi-step goal decomposition |
| Multi-Agent | Collaboration between multiple agents |
Memory & Learning (Part Two)
| Topic | Description |
|---|---|
| Memory Management | Short-term and long-term agent memory |
| Learning & Adaptation | Feedback loops and model updates |
| Model Context Protocol (MCP) | Tool and service interoperability |
| Goal Setting & Monitoring | Tracking and adjusting agent objectives |
Reliability (Part Three)
- Exception handling and recovery
- Human-in-the-loop oversight
- Knowledge retrieval via RAG
Scaling (Part Four)
- Inter-agent communication (A2A)
- Resource-aware optimization
- Advanced reasoning techniques
- Guardrails, evaluation, prioritization, exploration
Appendices
- Advanced prompting techniques
- GUI-to-real-world agents
- Framework overviews (AgentSpace, CLI agents)
- Reasoning internals and coding agents
See also
Source: obsidian vault —
Wiki/Articles/Agentic Design Patterns.mdandNOTES IN CS/agentic design.md