Skip to content
Agentic Levels

Everything starts here.

GuestLocal progress only
PreferencesSign in
01Start with one taskBest first move for beginners.02Check your LevelMeasure where you are.03Score an AI resultFind the habit to practice first.04Return to Your WorkScores, links, and checkpoints.
Start here

Begin

HomeThe main entry point.New to AIStart with one useful task.
Know where you are

Measure

Check your LevelUse this after you have tried AI.Fluency ScoreScore an AI result you can review.
Build the habit

Learn

LevelsLessonsTracks
Find the reference

Library

PromptsReferenceResourcesCompare Tools
Turn it into work

Apply

Your Next MoveChoose what AI should change next.Tool SetupGet the tools ready.
Come back later

Return

Your WorkScores, links, and checkpoints.My PathContinue from your level.Updates
Site

Site

PricingAboutFAQ & FeedbackPreferences

© 2026 Fuentes Studio

Privacy·Terms
yourCouncil
Ready to help
✦

What do you want to understand?

Ask anything about what you're learning.

← Back
L7Free

Deep dive · Jump here anytime when you want more on this idea

Connecting Claude to your real data with MCP

Model Context Protocol turns Claude into a live tool, not a text box

After this, you'll be able to explain what MCP is, install one MCP server in Claude Code, and use it to retrieve live data instead of typing information in manually.

Before you start

Complete Compounding Engineering first; this lesson builds on connecting AI outputs to external data sources and pipelines.

The idea

You've typed a question into Claude about your own files and gotten 'I don't have access to that.' That's not a Claude limitation. It's a setup problem MCP solves. MCP (Model Context Protocol) is the open standard that lets Claude call external tools. An MCP server is a small program that exposes a tool (a filesystem, a calendar, a database) to Claude in a format Claude understands. You install the server, add it to your Claude config file, and Claude can now call it as a tool during your conversation.

Here is the before and after: without MCP, asking Claude 'What meetings do I have this afternoon?' returns 'I don't have access to your calendar.' With the Google Calendar MCP server installed and configured, the same question returns your actual scheduled meetings, pulled live from your account.

Now try it: install the filesystem MCP server (it lets Claude read files from a folder on your computer), add it to your Claude Code settings, and ask Claude to summarize a document you have locally. You do not paste anything. Claude reads the file directly. That shift from 'you paste content in' to 'Claude pulls data when needed' is what MCP makes possible.

Try it (15 min)

Watch out for

  • Changing the config without fully restarting Claude; MCP servers load at startup and a partial restart will not pick up the new config
  • Granting write access to a filesystem server before you understand what Claude will do with it; start with read-only access
  • Confusing MCP with Zapier or Make. Those are no-code automation platforms. MCP is a direct protocol for tool calls inside Claude conversations.
  • Installing multiple servers at once; add one, verify it works, then add the next

Paste this into Claude

I want to set up [name the MCP server, e.g., filesystem, Notion, or Google Calendar] for Claude Code. I need Claude to be able to [describe the specific data task, e.g., read files from a folder called projects, search my Notion workspace, or check my calendar for the next 7 days]. Walk me through: 1) how to install the server, 2) the exact JSON to add to my Claude Code config file, 3) how to verify the server connected successfully after restarting Claude, and 4) a specific test prompt I can run to confirm Claude is reading live data and not guessing. After setup I will paste the test output back to you so you can confirm it worked correctly.

Created by potrace 1.16, written by Peter Selinger 2001-2019 What good looks like

  • The MCP server appears in Claude's available tools when you start a new session after adding the config
  • Claude uses the tool automatically when you ask a question that requires live data, without you telling it to
  • The test prompt returns actual data from your system, not a generated approximation
  • You can read the JSON config entry you added and explain what each field does

When this breaks

  • Breaks when the Claude config JSON has a syntax error because the server silently fails to load and Claude falls back to guessing instead of using the tool
  • Breaks when the server requires credentials you have not provided because the tool call returns an auth error that looks identical to Claude's normal 'I cannot access that' response

Created by potrace 1.16, written by Peter Selinger 2001-2019 You can now

✓

Ask Claude a question that requires the MCP tool to answer. Claude will show a tool-use indicator before responding. Verify the data it returns matches what you see in the source system directly.

Key takeaways

MCP servers connect Claude to live data. Without them Claude guesses; with them Claude reads. The config change is small; the capability shift is significant.

  1. 1MCP is an open standard that lets Claude call external tools (filesystems, APIs, databases) during a conversation
  2. 2Without MCP, Claude only uses what you type; with MCP, it pulls live data on demand through tool calls
  3. 3MCP servers load at Claude startup; a config change requires a full restart before it takes effect
  4. 4Start with read-only access and one server, then expand after you have verified the first one works

Was this helpful?

← Back to lessons