This is the single-screen index to the whole catalog: 56 agentic architectural patterns in one searchable navigator. Filter by pattern type or by Agentic AI maturity level (1–6), switch between grouped and flat views, and open any pattern for its summary, when-to-use guidance, and maturity placement. It’s the fastest way to find the pattern that fits the problem in front of you.
On the count (55 vs 56). The book and its companion reference catalog and skills package present 55 patterns across seven chapters. This navigator adds one consolidated entry — Advanced Model Tuning Patterns (SFT, PEFT/LoRA, DPO), grouped under Advanced Adaptation & Self-Improvement — bringing the interactive total to 56. The extra entry is a superset, not a discrepancy: everything in the 55-pattern catalog is here, plus model tuning.
↗ Open the navigator full-screen — recommended for search and on mobile.
How the catalog is organised
The 56 patterns sort along two axes. The type axis groups them into seven concerns; the maturity axis places each on a 1–6 ladder of agentic sophistication — and many patterns span more than one level, appearing in a simpler and a more advanced form.
| # | Pattern type | Patterns |
|---|---|---|
| 1 | Agent-Level Capability | 5 |
| 2 | Multi-Agent Coordination & Topology | 14 |
| 3 | Explainability, Compliance & Governance | 4 |
| 4 | Robustness & Fault Tolerance | 16 |
| 5 | Human-Agent Interaction | 5 |
| 6 | System-Level Infrastructure | 4 |
| 7 | Advanced Adaptation & Self-Improvement | 8 |
As in the rest of this series, Robustness & Fault Tolerance is the largest category (16) — the recurring signal that the hard engineering in multi-agent systems is what happens when things fail, not when they succeed.
The maturity ladder
- Level 1 — Basic Agentic Systems: single-agent tool execution via fixed workflows and direct function calls.
- Level 2 — Dynamic Single-Agent Workflows: dynamic tool selection, basic state, and RAG grounding.
- Level 3 — Introspective: step-by-step reasoning, self-reflection, retries, and prompt mutation.
- Level 4 — Multi-Agent Systems: collaborating specialists, top-down orchestration, governance and auditability.
- Level 5 — Advanced Coordination: meta-agents, market topologies, sandboxing, bidding and consensus.
- Level 6 — Self-Correcting Ecosystems: decentralized consensus, negotiation, coevolved training and emergent adaptation.
For the fuller treatments, this catalog connects to the rest of the series: the 55-pattern reference catalog (full context / problem / solution / code), the installable Claude Agent Skills, and the Skills Portal & Simulator for copy-paste system-prompt contracts. New to the series? Begin with the reading guide.
