agent design patterns

For a new task, eg reviewing applicant resumes, first try to teach an agent to do it before hiring someone to do it." — Article demonstrates practical agent task learning through concrete example of r

[DIRECT] "You have spent 15 years conducting and analyzing user research for top tech companies. You have a talent for reading between the lines and identifying patterns that others miss." — Article d

This article examines five essential design patterns that transform LLMs into AI agents capable of reasoning, planning, taking actions, and adapting autonomously." — Article directly frames its core t

Autonomous agents are defined with specific roles, goals, and back stories to enable specialized expertise" — CrewAI demonstrates role-based agent design pattern with specialized expertise definition

CrewAI mainly focuses on role delegation. Each agent gets: A role, A goal, A specific responsibility" — Article explains CrewAI's role-based architecture with concrete example showing role assignment

每轮循环检查舰队健康状况：保持健康任务后台运行，将故障/异常推送至 Slack 或 PagerDuty" — Demonstrates fleet health monitoring with automated status checks, anomaly detection, and multi-channel alerting (Slack, PagerDuty)

Design agent roles like Supervisor, Researcher, and Action Handler" — Article demonstrates role-based agent design with concrete examples (Supervisor, Researcher, Action Handler), showing a specific d

an "Agent" is really just a logical loop that can tell a LLM what tools it has access to, execute the tools the LLM says it wants to be executed, and feed the results of tool execution to a LLM" — Art

Agent( role='City Selection Expert', goal='Select the best city based on weather, season, and prices', backstory= 'An expert in analyzing travel data to pick ideal dest

Non-deterministic layer: model planning, tool choice, re-planning, evaluation. Deterministic layer: tool execution, input validation, idempotency, side-effect control." — Article provides a novel arch

To build reliable agents, start with strong foundations: pair capable models with well-defined tools and clear, structured instructions." — Article establishes core design principles for agent constru

The Model Context Protocol represents a paradigm shift in how we build AI agents" — Article positions MCP as introducing fundamentally new patterns for agent architecture and design, extending traditi

OpenAI's practical guide recommends maximizing a single agent's capabilities first because one agent plus tools keeps complexity, evaluation, and maintenance more manageable." — Direct recommendation

Learners will explore the fundamental concepts of AI agents, including reasoning, action-taking, and the ReAct pattern, and gain hands-on experience building these agents from scratch" — Course explic

[direct] "You're going to need less scaffolding around your agent as the models get better." — Introduces principle that agent architecture should evolve with model capability, reducing structural con

LangChain legacy agents use a language model as a "reasoning engine" to decide a sequence of actions to take. They are provided with user input (e.g. queries) and relevant steps executed previously."

LangGraph enables graph‑based workflows, with nodes and conditional edges, supporting loops, branching, parallel execution, and retry logic" — Article details a specific architectural pattern for agen

How you can use Printing Press to give your agent access to almost any website or app" — Article extends agent design patterns by showing how agents can be given broad external system access via API d

The agent uses this data to reflect, re-plan, or execute the next step." — MCP architecture demonstrates a key agent design pattern where agents iteratively reason over structured data to adjust their

Role-based architecture: You explicitly define agents with distinct roles (e.g., a researcher, analyst, writer) which helps structure complex tasks." — Article provides concrete evidence that role-bas

[direct] "writer_agent = Agent(role="Content Writer", goal="Create engaging articles based on research insights", backstory="You are a professional tech writer who simplifies complex topics."). Now we

The framework divides agents into three attributes—role, goal, and backstory—ensuring a thorough understanding of each agent's function. This structured approach mitigates under-specification risk, im

Drawing on vendor architectures (Google ADK, Anthropic, LangChain), current academic work (ACE framework, Google DeepMind's intelligent delegation), enterprise research (Deloitte, 2026; KPMG, 2026), a

These smart assistants combine powerful language models with the ability to perform actions. The result: systems that can not only understand and communicate, but can actually control and optimize bus

Testing Narrow AI Agents" — Article validates through extensive testing (200 hours) that narrow, purpose-specific agents outperform broad generalist agents.

Building AI agents for production needs clear control over what the model sees and when it acts" — Emphasizes input/output control and action gating as foundational design principles for production-re

The Role-Goal-Backstory Framework The most powerful agents in CrewAI are built on a strong foundation of three key elements" — Introduces the Role-Goal-Backstory Framework as a foundational design pat

Everything you build in this course is grounded in real-world enterprise requirements. You'll learn how to structure agent workflows for long-term scalability, fault tolerance, and ease of maintenance

Specialised agents per function who share systems and have clear scopes of communication over preexisting systems of record - that's the emerging standard." — Article identifies the architectural patt

[DIRECT] "Design conversational agent teams & advanced orchestration patterns with AutoGen." — Article highlights AutoGen course teaching agentic design patterns with multi-agent teams, providing conc

Then one shot their workflows and accept the output as valid" — Article identifies one-shot validation without iteration as anti-pattern, contradicting proper agent design methodology that requires en

That makes individual skills easier to evaluate, and it makes the agents themselves easier to manage as the system grows." — Provides empirical rationale for why skill composition improves agent desig

agents that write and review SQL, tools that summarize data lineage" — Article provides specific examples of agent-based systems (SQL-writing agents, data lineage tools) as production systems where co

Anthropic's recent blog post on "Building Effective Agents" lays out the difference between "agents" and "workflows", and presents a number of common patterns for both. Here, we implement every workfl

Learn what AI Agents are, how they work, and how to build them" — Article provides foundational guidance on agent architecture, design approaches, and implementation patterns using established framewo

Core settings for agents: **Role:** Clearly define the specific role or function of the agent, such as researcher, writer, or coder. **Goal**: Provide clear, concise instructions to guide decision-mak

Design, Orchestrate, and Deploy" — Title explicitly covers design of autonomous AI agents as a core objective of the guide.

[DIRECT] "Each node behaves like a small sub-agent. Generating web search queries is one skill, curating results is another, and packing them into state is a third. Modeling these as skills would like

forked subagents can inherit the same context as the main agent" — Article describes a new agent architecture pattern (forked subagents) introduced by Anthropic, demonstrating a concrete implementatio

i don't need any complex mechanisms to get the model to perform long accurate tasks, if i ever need anything that resembles a plan, you'll find it in my gh issues, most of the time the plan is already

I personally think it will come for free as we make our single-threaded agents even better at communicating with humans. When this day comes, it will unlock much greater amounts of parallelism and eff

Both promise structured collaboration between multiple agents, but they approach the problem from opposite ends. LangChain gives you composable agent graphs that are flexible and fine-grained, but req

lets it click buttons, read console logs, fill forms, and run multi-tab workflows" — Claude's agent capabilities expand from text-only to include browser interaction: DOM manipulation, console access,

Loss of Specialization: Similar to how a well-designed microservice outperforms a monolith in its specific domain, an agent specialized in LinkedIn can better leverage its context window to handle the

being able to repeatably design the right agent architectures is super important to helping you solve real, hard business problems." — Article emphasizes importance of repeatable, principled agent arc

complex workflows map more naturally here. Things like parallel execution, conditional branching, and long-running async tasks work much better in event-driven patterns in ways that synchronous chain

[INFERRED] "one of the failures of Watson was that, as it was commercialized, the use of UIMA was diminished, in effect breaking the spine of what architecturally made Watson successful" — Grady Booch

AI agents are no longer just simple responders. They are evolving into autonomous systems that can plan tasks, use external tools, and complete multi-step workflows." — Establishes that modern autonom

treating agents as microservices" — Article explicitly describes agents deployed with microservices architectural pattern for modularity and scalability

One approach to structuring generative AI applications is to divide them into distinct entities known as agents. Each agent is usually assigned specific roles and tasks, allowing them to handle partic