CC - Model Context Protocol (MCP)

1 Installing an MCP Server
- 1.1 Starting and Authenticating
2 Connecting to Atlassian
3 Using an MCP Tool
- 3.1 A Note on Security
4 Next Steps

By default, Claude Code comes with a set of built-in tools. Model Context Protocol, or MCP, is how you add more tools to that toolbox. It's a standard interface that lets you connect Claude Code to pretty much anything — external APIs, databases, cloud services, third-party platforms, whatever you can think of. In theory, MCP could connect your LLM to an API that controls a robot, not that you would ever, ever want to do that.

Installing an MCP Server

There's no built-in marketplace — you find MCP servers on GitHub, the Anthropic MCP docs, or through the claude.ai integrations list. Once you have one, you add it via the CLI.

For a local (stdio) MCP server — one you run as a process on your machine:

claude mcp add <name> -- <command> [args...]

Example with the Figma MCP:

claude mcp add figma -- npx -y figma-developer-mcp --figma-api-key=YOUR_KEY

For a remote (SSE/HTTP) MCP server — one hosted at a URL:

claude mcp add --transport sse <name> <url>

Example with the Atlassian remote MCP:

claude mcp add --transport sse atlassian https://mcp.atlassian.com/v1/mcp

Scope options control where the configuration is saved:

Flag	Config file	Use when

Flag	Config file	Use when
`--scope user` (default)	`~/.claude/settings.json`	You want this server available in all projects
`--scope project`	`.claude/settings.json`	You want to share the config with your team via source control
`--scope local`	`.claude/settings.local.json`	Project-specific but not committed to source control

Other useful MCP commands:

claude mcp list          # see all configured servers
claude mcp get <name>    # see details for a specific server
claude mcp remove <name> # remove a server

Ok, I’m glad you learned the hard way. If you don’t remember this syntax, remember you have an AI coding buddy with you. Just ask Claude to add the MCP server.

Starting and Authenticating

Once a server is configured, start a Claude Code session and run /mcp to see connection status and trigger any authentication flows. Some servers connect automatically; others (like Atlassian) will walk you through an OAuth flow in the browser.

Once connected, Claude Code automatically has access to all the tools that server exposes — no extra steps needed.

One thing worth doing: look through the new tools and disable anything you don't need. Some MCP servers expose a lot of capabilities, and you may not want Claude Code to have access to them for every project.

To disable an MCP server, run /mcp and select the server, hit [enter] and then choose “Disable”.

Connecting to Atlassian

Let’s try adding the Atlassian MCP server, which will give us access to Jira and Confluence. Run the prompt below.

✏️ Run this prompt

Add the Atlassian MCP server found here:
https://github.com/atlassian/atlassian-mcp-server

Once you are connected, use /mcp to check on the status. You will need to authenticate, so select the Atlassian connection and select “Authenticate”. A browser window will appear, follow the instructions to complete authentication.

Once your connection is established, let’s practice a real world scenario of reading a Jira ticket to build context for a new plan.

✏️ Run this prompt

Use the Atlassian MCP to look up info about:
https://bitovi.atlassian.net/browse/CAR-58

Using an MCP Tool

Using an MCP tool works just like using any built-in tool. Just describe what you want and Claude will figure out which tool to call.

For example, with the Figma MCP installed, you can paste in a link to a Figma design and ask Claude to read it. Claude will call the Figma MCP, get the design info, and then reason about it and write code based on it — same idea as when we used web search to look up the color palette, just now we have a Figma-specific tool that knows how to talk to Figma properly.

A Note on Security

MCP servers can expose some pretty powerful tools — and it's worth being thoughtful about what you give the agent access to.

Take an MCP server for a project management platform as an example. It might expose tools that let the agent create or edit issues, leave comments, or access sensitive project data — all authenticated under your personal account. Claude doesn't necessarily have full context about the downstream effects of using a tool; it'll just use whatever's available when it thinks it's appropriate.

On authentication tokens: some MCP servers require an API key or personal access token. Don't paste these directly into settings.json — especially if that file is committed to source control. Pass secrets as environment variables using the env block in your MCP config instead:

{
  "mcpServers": {
    "my-server": {
      "command": "npx",
      "args": ["-y", "some-mcp-server"],
      "env": {
        "API_KEY": "${MY_API_KEY}"
      }
    }
  }
}

Then set MY_API_KEY in your shell environment rather than hardcoding the value.

Next Steps

Now it's your turn — in the final exercise, you'll install and configure an MCP server of your choice and use it in a real session.

CC - Multi-Agent Workflows