Model Context Protocol (MCP)

An open standard for connecting AI models with external tools, data, and services. MCP solves complex integration challenges, making AI development faster, more efficient, and more secure.

Solving the Integration Problem: M×N → M+N

Before MCP: The M x N Problem

Connecting M AI models to N tools requires M x N custom integrations.

After MCP: The M + N Solution

If each model and tool adheres to the MCP standard, only M+N implementations are needed.

Core Concepts

MCP defines clear communication rules based on a client-server architecture.

Client-Server Architecture

Hover over a component to learn more.

🏠

MCP Host

↔
🛡️

MCP Client

↔
⚙️

MCP Server

Learn about the role of each component.

Communication Primitives

Clients and servers communicate using these fundamental message types.

Prompts (Server-Side)

Pre-canned instructions or templates to guide the AI model.

Resources (Server-Side)

Structured data to enrich the model's context (e.g., documents, code snippets).

Tools (Server-Side)

Actionable functions the model can invoke (e.g., sending a message).

Roots (Client-Side)

Entry points into the host's file system or environment that the server can access with permission.

Sampling (Client-Side)

A mechanism for the server to request the host AI to complete a task based on a prompt.

A Rapidly Growing Ecosystem

With participation from major AI companies, MCP is becoming an industry standard.

Key Participants

Even competitors like Anthropic, OpenAI, and Google DeepMind are adopting MCP, building an open ecosystem together.

Anthropic
OpenAI
Google DeepMind
Microsoft
Block
Replit
Sourcegraph

MCP Server Growth Trend

MCP Practical Guides

Learn how to connect MCP with Notion and Slack to extend your AI assistant's capabilities.

Notion: Connect Your Workspace to AI

Let your AI read, write, and search your Notion databases and pages.

Key Features

  • Query and search Notion databases
  • Create new pages and update existing ones
  • Summarize long documents and extract information
  • Automate task management and data entry

Setup Method

  1. Create an 'Internal Integration' in Notion and get the secret token
  2. Share the databases/pages you want to access with the integration
  3. Add the server configuration to your MCP client (e.g., Claude Desktop)

// Claude Desktop config example (claude-desktop-config.json)

{
  "mcpServers": {
    "notionApi": {
      "command": "npx",
      "args": [
        "-y", 
        "@notionhq/notion-mcp-server"
      ],
      "env": {
        "OPENAPI_MCP_HEADERS": "{\"Authorization\": \"Bearer YOUR_TOKEN_HERE\", \"Notion-Version\": \"2022-06-28\"}"
      }
    }
  }
}

MCP Security Center

With great power comes great responsibility. Review key security threats and mitigations.

Click a card to see the mitigation strategy.

The Future of AI Applications with MCP

MCP is more than just a technical standard; it's a key to fundamentally changing how AI interacts with our world. Based on standardization, openness, and robust security, we will enter an era of more intelligent, useful, and safe AI.

Reduced Development Costs

Eliminate redundant development and reduce maintenance overhead with M+N integration.

Enhanced Interoperability

Freely combine various models and tools without vendor lock-in.

Powerful AI Experiences

Implement more accurate and context-aware AI using real-time data and external tools.

Accelerated Innovation

Anyone can create and share new AI tools and services on top of an open standard.