Tutorials · Chapter C (3/4) · ~8 min
MCP — a standard plug for AI tools
Play → read → next
MCP is a standard way for AI apps to connect to tools and data — like USB for AI capabilities.
Playground
MCP: standard plugs for AI tools
Plug capabilities into an AI client. Ask needs Docs — only works if Docs is connected.
Recap
What you just did
You compared messy custom cables with MCP plugs. The ask that needed Docs only worked when Docs was connected — the idea is: capability in, capability available.
Teach
How it works
Think of three roles:
- Host / client — the AI app the human talks to
- MCP servers — small services that expose tools or data (files, docs, git…)
- Standard messages — how they advertise what they can do and how to call them
You still need permissions and good prompts. MCP doesn’t replace prompting or judgment — it standardizes connection.
Use it
When you'd use this
- Plug the same Docs capability into more than one AI client
- Let teams share tool servers instead of rewriting integrations
- Understand product demos that say “MCP support”
Watch out
Watch out
A standard plug isn’t automatic safety. You still decide what to connect, who can approve actions, and what data leaves the building. Also: MCP is not “a brand of LLM.”
Try next
Try this next
Name one tool you’d want as a reusable plug (calendar, notes, CRM). What would “connected” vs “unplugged” mean for a chat ask?