Choose frameworks and protocols deliberately
More agents means more coordination
A dinner assistant may need a few predictable steps, not a team of agents debating dinner. Start by asking what decision genuinely needs model judgment. Keep price arithmetic, eligibility, and permission enforcement in ordinary code.
LangChain supplies components for model applications; LangGraph focuses on stateful orchestration. Evaluate a framework against the workflow’s needs: persistence, inspection, recovery, and maintainability. A framework name is not an architecture explanation.
Check your understanding
Which task should stay deterministic in this prototype?
MCP and A2A solve different connection problems
Model Context Protocol (MCP) standardizes how applications connect to capabilities such as tools and resources. Agent2Agent (A2A) supports communication and task collaboration between agents. Neither automatically grants access to a restaurant platform.
Think of MCP as a common connector shape and A2A as a way for independent agent systems to exchange work. Read the MCP architecture overview and A2A overview for the protocol boundaries.
Check your understanding
Match each component to its role.
Write an architecture decision you can defend
For our first prototype: a small workflow, narrow tools, explicit state, and no live checkout connection. Add multiple agents only if a measured problem benefits from independent specialized work enough to justify extra messages, latency, and failure paths.
In your checkpoint, explain one component you would add next and what evidence would justify it. “Everyone uses it” is not evidence. A recurring recovery problem or a requirement for durable resumable tasks is a concrete reason to evaluate orchestration support.
Check your understanding
Would you add multiple agents to this dinner prototype now? Explain a criterion.
Build checkpoint
Open the Workflow Lab and follow the practice above. Each lab explains its inputs, its actual logic, and its limits. Save your observations below; your check-ins and reflection remain in Product Academy.
For an AI-assisted code change, download the lab bundle from the lab page. Ask your coding assistant to explain the relevant function first, change one behavior, and add one test that could fail. Compare the result with your prediction.
Save your experiment and check-in
Record your prediction, what you tried, what happened, and the evidence for your conclusion. Name one thing you can now do independently and one uncertainty to revisit.
Key takeaways
- Keep deterministic guarantees in application code.
- Choose protocols for a real integration need.
- Add orchestration complexity only when evidence supports it.