Home/Connecting Tools to Claude Code/Episode 1
IntermediateEpisode 1 of 410 min

CLI and MCP Tools for Claude Code

Understand how Claude Code can use external tools to read and write directly to your apps.

What Are CLI and MCP Tools?

So far, you've worked with files: export from an app, process with Claude Code, maybe import back. That works, but it's a lot of exporting and importing.

What if Claude could talk directly to your apps? Read your Google Sheets without exporting. Create Linear issues without opening Linear. That's what CLI tools and MCP servers enable.

The two connection methods

CLI tools (Command Line Interface) are programs that let you control an app from the terminal. Instead of clicking through Google Drive's website, you run gog drive list and see your files. Claude Code can use these same commands.

MCP servers (Model Context Protocol) are a newer standard. An MCP server sits between Claude and an app, translating requests. Claude says "create an issue in Linear" and the MCP server handles the API details.

Both achieve the same goal: Claude working directly with your apps instead of just files.

Why this matters

Without direct connections:

  1. You export data from App A
  2. Claude processes it
  3. You import results to App B
  4. Repeat whenever data changes

With direct connections:

  1. Claude reads from App A, processes, writes to App B
  2. Done

The export/import dance disappears. Your automations become truly end-to-end.

CLI tools: the basics

A CLI tool is a program you run from the terminal. Some examples:

  • gog — Google Workspace (Sheets, Docs, Drive, Gmail)
  • gh — GitHub (repos, issues, pull requests)
  • aws — Amazon Web Services
  • stripe — Stripe payments

When you install a CLI tool, you can do things like:

bash
# List Google Drive files
gog drive list

# Search Gmail
gog gmail search "from:[email protected]"

# Create a GitHub issue
gh issue create --title "Bug report" --body "Description here"

Claude Code can run these same commands. So instead of telling Claude "read this CSV I exported from Google Sheets," you can say "read the data from my Google Sheet called Q1 Budget."

MCP servers: the basics

MCP (Model Context Protocol) is a standard way for AI assistants to connect to external services. An MCP server is a small program that:

  1. Listens for requests from Claude
  2. Translates them into API calls
  3. Returns the results to Claude

The advantage: Claude doesn't need to know the details of each API. It just says what it wants, and the MCP server figures out how to do it.

MCP servers exist for:

  • Linear (project management)
  • Notion (docs and databases)
  • Slack (messaging)
  • Zapier (connects to 5000+ apps)
  • Many more

How Claude uses these tools

When you install a CLI tool or MCP server and configure Claude to use it, Claude gains new capabilities. You can then say:

Read my Google Sheet called "Q1 Budget" and summarize the totals by department.

Claude uses the CLI tool to access the sheet directly — no export needed.

Or:

Create a Linear issue for the bug we discussed.
Title: "Login button not working on mobile"
Assign to: @sarah

Claude uses the MCP server to create the issue in Linear.

What you need to set up

Each tool requires some setup:

For CLI tools:

  1. Install the tool (usually one command)
  2. Authenticate (log in with your account)
  3. Test that it works manually
  4. Claude can then use it

For MCP servers:

  1. Install the server
  2. Configure connection details (API keys, etc.)
  3. Tell Claude Code about the server
  4. Claude can then use it

The setup is usually a one-time thing. Once connected, Claude can use the tool whenever needed.

Security considerations

When you give Claude access to external tools, it can:

  • Read data from those services
  • Create, modify, or delete things
  • Take actions on your behalf

This is powerful but requires trust. Best practices:

  • Start with read-only access — Let Claude read but not write until you're confident
  • Use separate accounts for testing — Don't connect your production CRM on day one
  • Review what Claude does — Check the actions it takes, especially at first
  • Revoke access you don't need — If you stop using a connection, remove it

What's coming in this series

In the next lessons, you'll set up real connections:

  1. Google Workspace CLI — Read and write to Sheets, Docs, and Drive
  2. Linear MCP — Manage issues and projects directly
  3. Zapier MCP — Connect to thousands of apps through one integration

By the end, your Claude Code setup will talk directly to the tools you use every day.

Do I need to be technical?

You need to be comfortable with:

  • Running commands in the terminal (you've done this already)
  • Following installation instructions
  • Copying API keys into config files

You don't need to:

  • Write code
  • Understand how APIs work internally
  • Debug complex technical issues

If you've made it this far in the course, you have the skills. Let's connect some tools.

Next: Connect Google Workspace to Claude Code