Fundamentals of Generative AI
Agentic AI Concepts
Agentic AI systems use a foundation model to plan steps, call tools, keep memory, and coordinate with other agents to complete goals. For AIF-C01, know how agents differ from fixed workflows, what Model Context Protocol (MCP) does, multi-agent patterns, memory management, and context engineering.
Concepts
- An AI agent uses a foundation model to reason about a goal, decide which tools to call, and act on the results, instead of following one fixed sequence of steps.
- Model Context Protocol (MCP) is an open protocol that standardizes how agents connect to external tools and data sources.
- Multi-agent systems split complex work across specialized agents, for example a supervisor agent that routes tasks to collaborator agents.
- Agent memory includes short-term memory within a session and long-term memory that persists across sessions.
- Context engineering manages what goes into the context window, such as instructions, tools, retrieved content, and summarized history, to balance quality, cost, and latency.
Exam tips
- MCP is the standard way to expose tools to many compatible agents; Amazon Bedrock AgentCore Gateway can turn APIs into MCP tools.
- A supervisor with specialized collaborators fits work that spans several domains with many tools.
- Summarize or window long conversation history instead of resending every turn.