A reference catalog of agentic AI design patterns for software engineers building with LLMs. The live site is detached-node.dev. This repo is the source.
src/
├── app/
│ ├── (frontend)/ # Public site
│ │ ├── agentic-design-patterns/ # 24-pattern catalog (hub + satellites)
│ │ ├── posts/ # Essays
│ │ └── about/
│ └── (payload)/ # Payload CMS admin
├── data/agentic-design-patterns/ # Pattern source-of-truth (one TS file per pattern)
├── lib/schema/ # JSON-LD generators (BlogPosting, TechArticle, Person, etc.)
├── collections/ # Payload CMS collection definitions
└── components/ # Site UI
docs/ # Project docs (design system, deployment, SEO strategy)
PROJECT_BRIEF.md # Goals, non-goals, phase roadmap
CONTENT_MODEL.md # Payload collection field definitions
AGENTS.md # Conventions for AI-assisted contributors
SITEMAP.md # Public route inventory
- Prompt Chaining — Decomposes a task into a fixed chain of LLM calls, each consuming the previous output.
- Routing — Classifies the input, then forwards it to the matching prompt, model, or sub-agent.
- Parallelization — Fires N LLM calls in parallel and reduces them with a deterministic aggregator.
- Planning — Drafts a multi-step plan, executes it, and revises the tail when a step fails.
- Tool Use / ReAct — Loops a thought, a tool call, and the result until the agent emits a final answer.
- Code Agent — Wires the tool-use loop to a codebase, an editor, a shell, and a test runner.
- Evaluator-Optimizer — Drafts an answer, scores it against a rubric, and regenerates with the critique appended.
- RAG — Pulls documents from an external store and uses them as context for the model's answer.
- Agentic RAG — Lets the model issue retrieval as a tool call and decide when to stop searching.
- Reflexion — Writes a critique after each attempt and retrieves prior critiques on the next attempt.
- Orchestrator-Workers — Has a central orchestrator dispatch sub-tasks to fresh workers and merge their outputs.
- Multi-Agent Debate — Runs N agents on the same prompt, exposes them to each other's answers, and votes.
- Handoffs / Swarm — Lets one agent own the dialogue and hand control off via a transfer tool call.
- Guardrails — Wraps the model in input and output checks that block, rewrite, or refuse the response.
- Human in the Loop — Pauses at designated steps and resumes only after a person approves, edits, or rejects.
- Evaluation (LLM-as-Judge) — Scores model outputs with a stronger LLM applying a written rubric.
- Memory Management — Splits agent state into a working window plus episodic and semantic durable stores.
- Context Engineering — Decides which tokens fill the model's window, in what order, at what fraction of budget.
- Checkpointing — Persists run state at step boundaries so a fresh worker resumes after a crash.
- MCP — Standardises how an agent host discovers and calls tools across vendors over JSON-RPC.
- A2A — Lets one agent call another over HTTP by fetching an Agent Card and posting a Task.
- Streaming — Sends partial output frame by frame as it is generated, not after the model finishes.
- 12-Factor Agent — Catalogues twelve principles a team applies to push an agent from demo to production.
- Identity-Separated Review — Routes PR review through a separate machine-user identity on a different model tier.
- Agentic Patterns in Your Coding Workflow (May 15, 2026)
- Subagent Orchestration Workflow (Apr 24, 2026)
- In Defense of Tickets and Pull Requests (Apr 20, 2026)
- Rethinking Systems in the Agentic Age (Apr 19, 2026)
Next.js 16, React 19, TypeScript, Tailwind CSS 4, Payload CMS (Postgres backend), Google Cloud Run.
See:
PROJECT_BRIEF.md— phase roadmap, goals, non-goalsCONTENT_MODEL.md— Payload collection field definitionsAGENTS.md— conventions for AI-assisted contributorsSITEMAP.md— public route inventory
pnpm dev # localhost:3000
pnpm build # production build
pnpm lint # eslint + lint:adp
pnpm test:unit # vitest
pnpm test:e2e # playwrightMIT — see LICENSE.