AI itself has an ecosystem of tools and terms. Even experienced AI developers mix these four terms up constantly. Before we go any further, let's sort them out — because if you misclassify a tool, your expectations misfire in every direction.
🎯
Your goal: Drag each concept card to its correct role and analogy below.
Every concept has a job title (what it does) and a real-world analogy (what it feels like).
Get them all matched and the bigger picture clicks into place.
Match the Concept to Its Role
Drag each card from the left column into the correct slot on the right.
Concepts
🧠Agent
📡MCP
💻CLI
📋Skills
Roles & Analogies
Role: Sets goals, plans, decides, executes
Analogy: The brain of the operation. Understands the goal and decides what to do next.
Drop a concept here…
Role: Standard protocol connecting AI to external services
Analogy: A telephone operator. Routes calls between your agent and the outside world through a standardized exchange.
Drop a concept here…
Role: Runs commands directly on the system
Analogy: The hands. Direct, local, no middleman — like walking over in person instead of calling.
Drop a concept here…
Role: Packaged instructions for how an agent should behave
Analogy: A how-to manual. The agent reads it once and knows how to handle a whole category of tasks.
Drop a concept here…
Score: 0/4
Here's how they work together
🧠
Agent
understands the goal
→
📡
MCP
routes the call
→
💻
CLI
runs it locally
or
🌐
Service
via MCP server
→
✅
Result
back to the agent
MCP and CLI aren't rivals — they're two different routes the agent can take depending on the destination. MCP wins when there's no CLI alternative: Slack, Notion, Figma, Google Workspace. CLI wins when you need direct, local, debuggable access. The agent decides which road to take.