Agentic AI Notebook
0What is an AI Agent?1Why LLMs Need Agents2Anatomy of an Agent3Agent Lifecycle4Core Concepts5Agent Capabilities6Types of Agents7Agent Architectures8Agent Terminology9Current Agent Landscape10Planning11Reflection12Multi Tool13Self Correction14Build First AI Agent
Agent Foundations
Phase 4Module 10 of 15

Current Agent Landscape

Knowing the landscape helps pick build vs buy and stay interoperable.

Early cloud — AWS Azure GCP coexist; MCP is containers for tools.

Visual Workflows

Start here — scroll inside each diagram frame to explore, then use + / − to zoom up to 200% if needed.

100%
Loading diagram...

Scroll inside the frame to explore · use + / − to zoom up to 200%

MCP Tool Integration

100%
Loading diagram...

Scroll inside the frame to explore · use + / − to zoom up to 200%

How MCP standardizes tool access across agents and IDEs.

Key Takeaways

  • 1.IDE agents: Cursor Devin — code and terminal.
  • 2.Enterprise: Copilot Agentspace custom platforms.
  • 3.Frameworks: LangGraph CrewAI OpenAI SDK.
  • 4.Protocols: MCP for tools A2A for agents.

Real Example

Scenario

Series A startup: Cursor for engineering, LangGraph + custom tools for support, MCP servers for Salesforce and Zendesk instead of two bespoke OAuth integrations.

What you would do

Build vs buy: buy Cursor (IDE agent), build the support workflow (domain prompts + evals). MCP replaces custom adapters with standard tool servers. Monitor: MCP server uptime and time-to-add a new integration.

Practice Task

For your org, decide build vs buy for coding, customer support, and internal search. Name one framework and one protocol you would standardize on and why.

Code Walkthrough

Highlighted lines show where Current Agent Landscape happens in the code.

Current Agent Landscape
1# MCP client — list tools from a CRM server (landscape: protocol over custom SDKs)2from mcp import ClientSession  # import dependencies3
4async def list_crm_tools():5    async with ClientSession("salesforce-mcp") as session:6        tools = await session.list_tools()  # key line for Current Agent Landscape7        return [t.name for t in tools.tools]  # return the result8
9# Agent picks from standardized tool list instead of hard-coded REST wrappers

Cheat Sheet

Quick recap

quick ref
  • •MCP = standard tool layer
  • •LangGraph = graph orchestration
  • •Eval before every deploy
  • •Build vs buy per use case
  • •Watch protocol standards

Common Mistakes

  • ✕Skipping evaluation for Current Agent Landscape before production
  • ✕No logging or tracing around current agent landscape steps
  • ✕Ignoring cost and latency implications