Descriptions:
This official tutorial from the Claude channel (Anthropic) provides a concise but thorough introduction to using Model Context Protocol (MCP) inside Claude Code. MCP is an open standard that connects Claude Code to external tools and data sources — bridging the gap between the AI assistant and the systems where real development work lives, including databases, project management platforms like Linear, and documentation sources like the Context7 server for up-to-date dependency docs.
The video explains the two server types: HTTP servers for remote services hosted by the provider and connected over the network, and STDIO servers for local processes running directly on the developer’s machine. Servers can be scoped at three levels — local (current project only), user (all projects), or project-wide via a `.mcp.json` file checked into version control, which automatically distributes the same MCP configuration to every contributor on a codebase. The `/mcp` slash command inside a Claude Code session shows connection status and lets users disable servers on the fly.
A practical note emphasized throughout: every configured MCP server adds tool definitions to the context window even when idle, eating into available context. If total MCP tool definitions exceed 10% of the context window, Claude Code automatically switches to on-demand tool search mode, which may underperform. For tools with CLI equivalents — such as `gh` for GitHub or `aws` for the AWS CLI — using the CLI directly is recommended as more context-efficient than running a persistent MCP server.
📺 Source: Claude · Published May 09, 2026
🏷️ Format: Tutorial Demo
