Frequently Asked Questions

MCP (Model Context Protocol) is an open standard that lets AI models connect to external data sources and tools through a unified interface. It acts as a universal adapter between AI assistants and the services they need to access. Learn more in our introductory blog post.

MCP uses a client-server architecture where an AI host (like Claude Desktop) connects to MCP servers that expose tools, resources, and prompts. The client sends JSON-RPC messages to servers, which execute actions and return results. Check our tutorials for hands-on walkthroughs.

Yes, MCP is completely free and open source. The protocol specification, SDKs, and many community-built servers are available at no cost. You can browse free servers in our server directory.

MCP was created by Anthropic, the company behind Claude. It was released as an open standard to enable interoperability across AI platforms and tools. Read more about the ecosystem in our blog.

Traditional APIs require custom integration code for each service, while MCP provides a single standardized protocol that works across all compatible tools. This means one MCP server can be used by any MCP-compatible client without additional coding. See our explainer for a deeper comparison.

Yes, the MCP specification and reference implementations are open source. Anthropic maintains the core SDKs, and the community contributes servers and tooling. Explore community servers in our directory.

Claude (via Claude Desktop and Claude Code) has native MCP support. Other clients like Cursor, Windsurf, and Continue also support MCP. Check our clients page for the full list of supported AI tools.

The MCP SDK is a set of libraries that help developers build MCP servers and clients. SDKs are available for TypeScript, Python, and other languages. Our tutorials section includes guides for getting started with each SDK.

Most MCP servers can be installed via npm or pip with a single command. You then add the server configuration to your client's config file. Follow our installation tutorial for step-by-step instructions.

You need an MCP-compatible client (like Claude Desktop or Cursor) and at least one MCP server. No coding experience is required for basic setup. Visit our clients page to choose your starting point.

The Filesystem server is a great starting point since it lets your AI read and write local files. The Brave Search server is another popular first choice for web search capabilities. Browse our server directory to find what fits your needs.

Open Claude Desktop settings, navigate to the Developer section, and edit the MCP configuration file to add your server definitions. Each server needs a command and optional arguments. See our Claude Desktop setup guide for details.

In Cursor, go to Settings > MCP and add your server configurations. Cursor supports both stdio and SSE transports. Check our Cursor setup guide for a complete walkthrough.

No, you can use pre-built MCP servers without writing any code. Installation typically involves running one command and editing a JSON config file. Our tutorials walk you through the entire process.

You can run as many MCP servers as your system resources allow. Most clients support multiple simultaneous server connections with no hard limit. Explore the variety of available servers in our directory.

An MCP server is a program that exposes tools, resources, or prompts to AI clients through the MCP protocol. It acts as a bridge between the AI and external services like databases, APIs, or file systems. Browse available servers at our server directory.

The MCP ecosystem is growing rapidly with hundreds of community-built servers. MCPgee catalogs 60+ of the most popular and well-maintained servers across various categories. Check our directory for the latest count.

Official and well-maintained community servers are generally safe, but you should always review what permissions a server requests. MCP includes a consent mechanism so the AI must ask before executing sensitive actions. Read our security best practices guide.

Yes, you can build custom MCP servers using the official SDKs. The TypeScript and Python SDKs make it straightforward to expose your own tools and data sources. Follow our Python server tutorial to get started.

Official SDKs exist for TypeScript and Python, which are the most popular choices. Community SDKs are also available for Rust, Go, Java, and other languages. Explore language options in our tutorials.

Compare servers based on maintenance activity, feature set, documentation quality, and community adoption. MCPgee provides detailed profiles for each server to help you decide. Use our server directory filters to narrow your search.

Filesystem, Brave Search, PostgreSQL, GitHub, and Puppeteer are among the most widely used MCP servers. Popularity varies by use case - developers often start with code-related servers. See the top-rated servers in our directory.

It can be safe with proper precautions. Use read-only database credentials, restrict access to specific tables, and always review the queries your AI generates before execution. Our security guide covers database safety in detail.

Store credentials in environment variables rather than config files. Use secrets managers for production setups and rotate keys regularly. Check our tutorials for secure configuration examples.

Only if the server is specifically designed for internet access (like Brave Search or Fetch). Local-only servers like Filesystem operate entirely on your machine without network calls. Review server capabilities in our directory.

Most file-related MCP servers accept an allowed-directories configuration that limits which folders the AI can access. Always scope access to the minimum required directories. See our security best practices for configuration tips.

Community servers vary in quality and security. Look for servers with active maintenance, open-source code you can audit, and positive community feedback. MCPgee reviews servers listed in our directory for quality assurance.

Review the server's source code on GitHub, check what permissions it requests, and test it in a sandboxed environment first. Most well-maintained servers are fully open source. Our security guide includes an audit checklist.

Claude Desktop, Cursor, Windsurf, Continue, and VS Code (via Copilot Chat) all support MCP. The ecosystem is growing as more tools adopt the standard. See our clients page for the complete list.

ChatGPT does not natively support MCP at this time. MCP is currently best supported by Claude and code editors like Cursor. Visit our clients page to see which tools have MCP support.

VS Code supports MCP through GitHub Copilot Chat in agent mode. You can configure MCP servers in your workspace settings. Read our VS Code setup guide for configuration details.

Yes, you can run multiple MCP servers simultaneously. The AI client can use tools from different servers in the same conversation, combining capabilities seamlessly. Browse complementary servers in our directory.

MCP supports stdio (standard input/output) for local servers and SSE (Server-Sent Events) for remote servers. Stdio is the most common for desktop usage. Learn about transport options in our tutorials.

Yes, MCP servers can run on remote machines using SSE transport over HTTP. This enables team-shared servers and cloud-hosted tool access. Our blog covers remote deployment patterns.

Resources are read-only data sources (like files or database records) that the AI can access, while tools are executable actions (like running a query or creating a file). Both are exposed through the same server. Learn the differences in our MCP overview.

MCP prompts are reusable prompt templates that servers can expose to clients. They allow servers to suggest specific workflows or interaction patterns to the AI. Explore servers with custom prompts in our directory.

With stdio transport, the client launches the server as a subprocess and communicates via standard input and output streams using JSON-RPC messages. This is the simplest and most common transport for local servers. See our tutorials for implementation details.

SSE (Server-Sent Events) transport allows MCP servers to run as HTTP services, enabling remote access and multi-client connections. It's ideal for shared or cloud-hosted servers. Learn about SSE setup in our tutorials.

Yes, MCP supports streaming responses for long-running operations. Servers can send progress notifications and partial results while processing. Check our blog for articles on advanced MCP patterns.

If a server crashes, the client detects the disconnection and can attempt to restart it automatically. The AI will inform you that the tool is temporarily unavailable. Most clients handle recovery gracefully - see our clients page for details on each client's behavior.