Everything about Claude Code, MCP protocol, semantic MCP server search, and our interactive quiz.
Claude Code is a CLI tool developed by Anthropic. It lets you use Claude directly in your terminal to code, debug, explore codebases, and automate development tasks. It's an AI agent that can read, write, and execute code in your project.
Install via npm: `npm install -g @anthropic-ai/claude-code`. Then run `claude` in a terminal to start a session. An Anthropic account with API access is required.
Claude Code itself is free and open-source. However, it uses the Claude API which is billed per usage (tokens sent/received). Anthropic also offers Claude Code through the Max subscription which includes a usage quota.
MCP is an open protocol created by Anthropic that allows Claude to connect to external tools (databases, APIs, services). It's like plugins: each MCP server exposes "tools" that Claude can call. There are over 4,700 MCP servers listed publicly.
Claude.ai is the classic conversational web interface. Claude Code is a command-line agent designed specifically for software development: it can read your files, execute commands, modify code, and navigate your project — directly in the terminal.
Yes, that's its primary use case. Just run `claude` in your project folder. Claude analyzes the codebase, understands the architecture, and can modify existing files, create branches, make commits, and even open pull requests.
Claude Code works with all programming languages — TypeScript, Python, Rust, Go, Java, etc. It can also work with configuration files, SQL, Markdown, and any text file type.
CLAUDE.md is a configuration file placed at the root of your project. It contains persistent instructions for Claude Code: code conventions, tech stack, useful commands, project context. Claude reads it automatically at the start of each session.
There are directories like Smithery.ai, Anthropic's official MCP Hub, or our semantic search engine at claudequiz.app/mcp-search. The latter lets you describe what you want to do in natural language and returns the most relevant MCP servers from 4,700+.
Use the command `/mcp add <name> -- <command>` in Claude Code. For example: `/mcp add github -- npx @modelcontextprotocol/server-github`. Configured servers persist and are available in every session.
As of April 2025, over 4,700 MCP servers are listed publicly, offering more than 32,000 tools in total. They cover diverse domains: databases, cloud APIs, e-commerce, productivity, web search, design, etc.
An MCP tool is a function exposed by an MCP server that Claude can call. For example, a GitHub server exposes tools like `create_issue`, `list_pull_requests`, or `search_code`. Each tool has an input schema (expected parameters) and returns a result.
It's a platform that combines an interactive quiz on Claude Code with a semantic search engine for MCP servers. The quiz helps learn Claude Code commands, shortcuts, and concepts. The MCP engine helps find the right tools for your projects.
The quiz contains over 225 questions across 6 categories: commands, keyboard shortcuts, concepts, MCP, workflows, and skills. New questions are generated automatically each week from official Anthropic documentation.
The engine uses vector embeddings to understand your natural language query. It compares your description (e.g., "manage Jira tickets") with descriptions of 4,700+ MCP servers and 32,000+ tools, then returns the most relevant results ranked by similarity score.
Our engine is semantic: instead of searching by exact keywords, it understands the intent behind your query. It also shows the specific tools that match, not just the server. And the AI deep analysis re-ranks results based on your specific context.
Yes, a weekly cron automatically generates new questions by analyzing official Anthropic documentation via AI. The quiz stays synchronized with the latest Claude Code features.
The homepage and FAQ are accessible without an account. To play the quiz and use the MCP search engine, you need to create a free account (via GitHub or Google). This lets you save your progress and favorite MCPs.
Yes, once signed in, you can bookmark your favorite MCP servers to easily find them later. Great for building your personalized MCP toolkit.
Try the MCP search engine: Search for an MCP server →
Here are 20 questions from the 225+ available. The full quiz covers many more topics and difficulty levels.
Enter — Unlike many editors that use Ctrl+Enter, Claude Code uses simple Enter to send. For a line break, use Shift+Enter or Option+Enter.
Escape — The Escape key immediately stops the current generation without quitting the session. Ctrl+C also works but may quit Claude Code if pressed twice.
/model — The /model command lets you switch Claude models (Opus, Sonnet, Haiku) mid-session without restarting.
/history — /history shows previous sessions. You can resume a past conversation with `claude --resume` or `claude --continue`.
/exit or Ctrl+C twice — The `/exit` command closes the session cleanly. Ctrl+C interrupts the current generation, and if pressed twice quickly, quits Claude Code.
Generate a CLAUDE.md file for the project — /init analyzes your codebase and automatically generates a CLAUDE.md file with detected conventions, useful commands, and project context.
Compress the conversation context — When the conversation gets long, /compact summarizes previous exchanges to free up context window space while keeping essential information.
claude -p "your prompt" — The -p flag (or --print) executes a single prompt without entering interactive mode. Useful for scripts and CI/CD pipelines.
claude --continue — --continue (or -c) resumes the most recent session with all its context. Handy when you quit accidentally or want to continue the next day.
Claude plans and executes multiple steps autonomously — In agentic mode, Claude doesn't just answer: it breaks down a complex task into steps, reads files, executes commands, and modifies code autonomously until the goal is reached.
Shell commands executed automatically in response to events — Hooks automate actions: for example, running a linter after each file modification, or running tests after a commit. They're configured in settings.
The maximum amount of text Claude can process in a conversation — The context window includes conversation history, files read, command results. When it approaches its limit, Claude Code automatically compresses older messages.
The level of autonomy granted to Claude for executing actions — Claude Code offers several modes: from "ask everything" (asks confirmation for everything) to "auto" (executes actions without asking). This controls whether Claude can read/write files or execute commands without your approval.
/mcp — The /mcp command without arguments shows all currently configured MCP servers with their status (connected/disconnected) and available tools.
In ~/.claude/settings.json or .claude/settings.json in the project — Global MCP servers are in ~/.claude/settings.json. Project-specific servers are in .claude/settings.json at the project root. Project scope takes priority.
Yes, if the server is designed for it — it's a separate process — Each MCP server is an independent process that can make HTTP requests, access APIs, read the filesystem, etc. Claude communicates with it via the MCP protocol (stdin/stdout or SSE).
Local = process on your machine, remote = server accessible via HTTP/SSE — Local servers (stdio) are launched by Claude Code on your machine. Remote servers (SSE/HTTP) are hosted somewhere and Claude connects to them via URL. Remote ones can be shared between multiple users.
It uses git directly with git add, commit, push commands — Claude Code executes git commands in your terminal. It can create branches, write descriptive commit messages, and push changes — always asking confirmation for destructive actions.
Yes, via the GitHub CLI (gh) or Git APIs — Claude Code uses `gh pr create` to create PRs directly from the terminal. It automatically generates a title and description based on the changes made.
With non-interactive mode: claude -p "instruction" — The -p flag runs Claude in one-shot mode, perfect for CI. Combined with --output-format json, it can be integrated into automated scripts for code review, docs generation, or migrations.
Want to test your knowledge? Start the interactive quiz →