DEV Track Spotlight: Building Agentic AI - Amazon Nova Act and Strands Agents in Practice (DEV327)
Source: Dev.to
🎥 Watch the Full Session
[Insert video embed or link here]
The Evolution of Agentic AI
Haowen opened by examining the evolution of generative AI, which follows a clear progression:
| Spectrum | Description |
|---|---|
| Low‑agency | Rule‑based systems that require high human oversight. |
| High‑agency | Systems that can operate independently and even make strategic decisions autonomously. |
“Although we are still in the very early stage of agentic AI, we have a vision of a future world powered by AI agents,” Haowen explained, referencing Sequoia Capital’s keynote at AI Ascent 2025.
Sequoia Capital’s Vision
- Future AI systems will evolve beyond tool‑like attributes into autonomously operating intelligent agents with reasoning, planning, collaboration, and high‑autonomous operation.
- Their 2030s outlook includes an “agent economy” that functions like a global neural network of interconnected agent operations.
Implications
- The first one‑person unicorn – a company created and operated by a single individual with a $1 billion valuation.
- New organizational models where a single person can run a business at the scale of a full startup.
Interoperability Protocols
Just as TCP/IP laid the foundation for the Internet, AI agents need standards for communication. Emerging protocols such as:
- Model Context Protocol (MCP)
- Agent‑to‑Agent (A2A)
AWS is actively contributing to both MCP and A2A, leveraging its decades of distributed‑systems expertise.
Demo 1 – Weather‑Info Agent (Amazon Nova Act)
Haowen demonstrated an AI agent that automatically retrieves local weather data from the Hong Kong Observatory website. The demo showed how Nova Act can act like a human web‑scraping engineer: locate information, capture verification screenshots, and return structured results.
How It Works
-
Forecast Function – Guides Nova Act on which pages to capture and what actions to perform.
-
Natural‑Language Task – Example prompt:
read and extract the complete nine‑day weather forecast information visible on this pagewith the Hong Kong Observatory URL as the starting point.
-
MCP Call – Uses the
@mcpdecorator; runs Nova Act in a separate thread. -
Portability – Changing the URL adapts the agent for any country.
Demo 2 – Mathematical Animations (Strands Agents SDK + Manim)
The second demo created a mathematical animation using Manim, a Python library for high‑quality visualizations. The Strands Agents SDK processed a user prompt, generated a precise Manim script, and produced a polished animation.
Minimal Code Overview
| Step | Description |
|---|---|
| Import Essential Components | from strands import Agent and from mcp import MCPClient |
| Establish MCP Connection | Set up input/output transports to the Manim MCP server |
| Prepare Tools | Query the MCP server for available tools and attach them to the agent |
| Process Natural‑Language Prompt | Example: “create a Manim animation that visualizes a cubic function graph from X = ‑3 to X = 3.” |
| Adaptive Problem Solving | When the agent detected environment incompatibilities, it generated a simplified version to complete the task. |
The entire educational video creation took just 30 minutes.
About Strands Agents SDK
- Open‑source framework created by AWS.
- Enables developers to build AI agents with minimal code.
- Handles planning, chain‑of‑thought, tool‑calling, and orchestration, letting you focus on unique business logic.
Building a truly functional AI agent often requires 80‑90 % effort on infrastructure rather than core logic. Strands Agents SDK abstracts that away.
Multi‑Agent Systems (Jacky Wu)
Jacky took the session deeper into multi‑agent architectures, explaining why they matter:
| Benefit | Explanation |
|---|---|
| Specialization | Domain‑specific agents increase precision and deliver superior outcomes when combined. |
| Scalability | Agents can be added or removed without impacting the whole workflow. |
| Maintainability | Smaller agents are easier to control, monitor, debug, and test (debugging 100 lines vs. 10 000 lines). |
| Cost Efficiency | Different agents can use models matched to task complexity, balancing speed, performance, and cost. |
“Multi‑agent systems are like a well‑coordinated team,” Jacky explained. “A monolithic …”
(The sentence was truncated in the original transcript.)
Takeaways
- Agentic AI is moving from low‑agency tools to high‑agency autonomous systems.
- Interoperability standards (MCP, A2A) are essential for a future “agent economy.”
- Amazon Nova Act and Strands Agents SDK let you prototype powerful agents with just a few lines of code.
- Multi‑agent architectures provide specialization, scalability, maintainability, and cost efficiency—key for production‑grade AI solutions.
Ready to start building?
Explore the Amazon Nova Act documentation and the Strands Agents SDK repository to create your own agentic applications today.
Multi‑Agent Collaboration Patterns in Strands Agents SDK
“An agent is like a trader managing a portfolio all by himself.
Multi‑agent is like a team where you have analysts doing market research, quants doing modeling, risk managers managing portfolio risk, and IT maintaining infrastructure. When they work together, they achieve superior performance.”
Strands Agents SDK supports four key multi‑agent collaboration patterns:
| Pattern | Description | Demo Highlights |
|---|---|---|
| Orchestrator | An orchestrator agent receives user requirements, understands intent, and calls specialized agents to perform specific tasks. The output of one agent becomes the input for the next agent in the chain. | Jacky showed a quantitative‑hedge‑fund use case where a quant‑research orchestrator called market‑data tools, back‑test tools, strategy‑generation agents, and results‑summary agents. The system returned P/L metrics, Sharpe ratios, buy/sell conditions, and improvement suggestions. |
| Swarm | No single orchestrator; agents communicate dynamically and share a common context (who is in the system, each agent’s inputs/outputs, historical interactions). | Jacky demonstrated a blog‑creation workflow with research, creative, critical, and summarize agents. The summarize agent requested more detail from the creative agent before producing a social‑media summary, illustrating adaptive hand‑off. |
| Graph | Agent relationships are defined as a graph: each agent is a node, edges dictate allowed communication and information flow. This gives predictability and fine‑grained control. | Jacky showcased an agentic fund manager (built by AWS colleague Kang Wan Lee) using Strands Agents SDK together with LangGraph. Nodes represented financial analysis, portfolio construction, and risk management; edges controlled data flow between them. |
| Workflow | A well‑defined sequence of tasks with explicit dependencies and priorities. Parallel execution is possible when tasks are independent. | Ideal for repeatable pipelines (e.g., ETL). The demo highlighted an automated data pipeline with extraction → transformation → loading steps. |
Key Takeaways from the Session
- Start with Clear Use Cases – Real‑world demos (weather agent, mathematics video creator) showed how quickly functional AI agents can be built.
- Leverage Open Source – Strands Agents SDK gives developers source‑level control while eliminating 80‑90 % of boilerplate code.
- Choose the Right Pattern – Match your solution’s requirements for control, flexibility, and scalability to one of the four patterns.
- Match Models to Tasks – Use different LLMs for different sub‑agents (e.g., Claude Sonnet for coding, Amazon Nova for summarization) to optimize speed, performance, and cost.
- Embrace the Stochastic Mindset – Success with agentic AI requires organizational and mental‑framework changes, not just technology.
- Protocols Matter – Just as TCP/IP enabled the internet, MCP and A2A protocols are laying the foundation for the future agent economy.
Haowen: “Whether you embrace it or not, the era of agentic AI has arrived.”
The tools and frameworks highlighted—Amazon Nova Act, Strands Agents SDK, AgentCore Gateway, and AgentCore Runtime—provide everything developers need to build production‑ready agentic AI applications today.
Practical Use‑Case Summary
| Pattern | Example |
|---|---|
| Agent as Tools | Customer‑service chatbot where an orchestrator calls technical‑account‑manager, billing, and summarize agents to deliver comprehensive responses. |
| Swarm | Multidisciplinary incident‑response team that dynamically collaborates across networking, database, and application layers. |
| Graph | Data‑validation pipeline with explicit error‑handling paths: separate agents for clean data vs. data requiring correction. |
| Workflow | Automated ETL pipeline: extraction → transformation → loading into a data warehouse. |
DEV Track Spotlight – AWS re:Invent 2025
This post is part of DEV Track Spotlight, a series highlighting standout sessions from the AWS re:Invent 2025 Developer Community (DEV) track.
- 60 unique sessions
- 93 speakers (AWS Heroes, Community Builders, User‑Group Leaders, AWS staff)
Topics Covered
- 🤖 GenAI & Agentic AI – Multi‑agent systems, Strands Agents SDK, Amazon Bedrock
- 🛠️ Developer Tools – Kiro, Kiro CLI, Amazon Q Developer, AI‑driven development
- 🔒 Security – AI‑agent security, container security, automated remediation
- 🏗️ Infrastructure – Serverless, containers, edge computing, observability
- ⚡ Modernization – (additional topics omitted for brevity)
- 📊 Data – Amazon Aurora DSQL, real‑time processing, vector databases
Each post in this series dives deep into one session, sharing key insights, practical takeaways, and links to the full recordings. Whether you attended re:Invent or are catching up remotely, these sessions represent the best of our developer community sharing real code, real demos, and real learnings.
Follow along as we spotlight these amazing sessions and celebrate the speakers who made the DEV track what it was!