MCP and Claude Code
Anthropic created MCP to standardize how context is provided to LLMs. Claude Code as an AI coding agent leverages MCP to enhance its capabilities. However, people think that Claude Code can only do things using MCP, which is not true. Claude Code has out-of-the-box tools that can do several things. In other words, Claude can do a lot without MCP.
I asked claude what tools it has inside of the box and this is what I got:
Prompt:
Tell me what tools Claude Code has inside-of-the-box. I don't want to know about MCP (external nor remote).
Result:
⏺ Claude Code has these built-in chat/interaction tools:
Task Management:
- TodoWrite - Create and manage task lists during coding sessions
User Interaction:
- AskUserQuestion - Ask questions during execution with multiple choice options (supports multi-select)
Extensibility:
- SlashCommand - Execute custom slash commands defined in.claude/commands/
- Skill - Execute skills (currently none installed in your environment)
Information Gathering:
- WebSearch - Search the web for current information
- WebFetch - Fetch and analyze content from URLs
Specialized Agents:
- Task - Launch specialized agents for complex tasks:
- Explore - Fast codebase exploration
- Plan - Planning agent
- general-purpose - Multi-step task handling
- changelog-generator - Generate changelogs from git commits
- marketing-agent - Translate technical content to plain language
All other tools (Bash, Read, Write, Edit, Glob, Grep, NotebookEdit) are file/system operation tools rather
than chat-specific features.
As you can see, Claude can do a lot without MCP. One important thing that I need to call out is that because Claude Code can run bash commands in your terminal, that's perhaps one of the most powerful things in the universe. MCP is great, but there are lots of bad actors out there and you need to be careful using MCP. Now, using Claude Code tools and only leveraging what's in your machine, you are more secure. Because if you have a vetted process to get tools on your machine or you do your due diligence for what you install, then you can piggyback on that.
Running bash is a Swiss Army knife. You can do almost anything with bash. For instance, here is a list of things, and not limited to what Claude can do with bash:
- Run any docker container: Same for podman. Yes, there is an attack vector here, but if you get your containers from an internal trusted registry, you are good.
- Execute: Claude can also execute any command on a container, so it can run SQL, it can deploy pods in K8s, it can create data, it can do anything you can do in bash.
- Script: Claude can create the script you need to get something done and executed.
Status Line
Claude status line is a feature where you can customize the status line that appears at the bottom of the Claude Code interface. You can do that to display useful information; there is an MCP for that called ccstatusline. You can use that MCP to create your own status line.
To install:
npx ccstatusline@latest
CCStatusline Configuration | v2.0.21
╭────────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ > Preview (ctrl+s to save configuration at any time) │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
Model: Claude | Ctx: 18.6k | ⎇ main | (+42,-10)
Main Menu
▶ 📝 Edit Lines
🎨 Edit Colors
⚡ Powerline Setup
💻 Terminal Options
🌐 Global Overrides
🔌 Uninstall from Claude Code
🚪 Exit
Configure any number of status lines with various widgets like model info, git status, and token usage
Then do all configs you want and just click on "Install on Claude Code".
Useful MCPs
Here is a list of useful MCPs for Claude Code, again, use with caution.