This documentation covers two main components:

1. AI Rules System (Project & Global Rules)
2. Model Context Protocol (MCP) Integration

Project rules provide path-specific configurations stored in .cursor/rules directory, offering granular control over AI behavior within your project. Key features include:

• Semantic Descriptions: Define specific conditions for rule application
• File Pattern Matching: Implement glob patterns for targeted file/folder control
• Automatic Attachment: Dynamic rule inclusion based on file references
• Reference Files: Utilize @file syntax to incorporate contextual information

Global rules are configured through:

• Path: Cursor Settings > General > Rules for AI
• Purpose: Define project-independent settings like:
  • Output language preferences
  • Response length parameters
  • Universal formatting rules

Legacy Support Note: While .cursorrules files remain supported for backward compatibility, migration to the new Project Rules system is recommended for enhanced functionality.

MCP serves as a standardized plugin system for Cursor, enabling:

• External tool integration
• Data source connections
• Enhanced LLM context management

• Connects Cursor with external systems
• Supports any programming language capable of stdout output or HTTP endpoint serving
• Facilitates quick implementation with flexible technology stack options

1. SSE Transport

   • Remote/local execution capability
   • User-managed operation
   • Network-based communication
   • Cross-machine sharing support
   • Configuration: URL to /sse endpoint

2. stdio Transport

   • Local machine execution
   • Automatic Cursor management
   • Direct stdout communication
   • Single-user access
   • Configuration: Shell command execution

{
  "mcpServers": {
    "server-name": {
      "command": "python",
      "args": ["mcp-server.py"],
      "env": {
        "API_KEY": "value"
      }
    }
  }
}

1. Global Configuration

   • Path: ~/.cursor/mcp.json
   • Scope: All Cursor workspaces
   • Use: Cross-project tool availability

2. Project Configuration

   • Path: .cursor/mcp.json
   • Scope: Project-specific
   • Use: Project-dedicated tools

• Automatic tool selection based on relevance
• Manual tool invocation through name/description
• Integrated response display in chat

• Default tool approval system
• Argument inspection capability
• Optional Yolo mode for automated execution
• Detailed tool call response tracking

1. Tool Quantity Restriction

   • Maximum 40 tools per Agent session
   • Applies to multiple active MCP servers

2. Remote Development Constraints

   • Limited functionality in SSH environments
   • Local machine communication requirement

3. Resource Support Status

   • Current focus on tool capabilities
   • Resource support planned for future releases

• Cursor AI Rules
• Cursor Directory
• Cursor Rules Reference
• MCP Documentation

• Enhanced remote development support
• Expanded resource capabilities
• Continued protocol optimization

This documentation maintains comprehensive coverage while providing:

• Clear hierarchical structure
• Improved readability
• Technical accuracy
• Practical implementation guidance