What is MCP (Model Context Protocol)?

Model Context Protocol (MCP) is an open standard that allows AI assistants to connect with external tools and data sources. Developed by Anthropic and released as an open specification, MCP provides a standardized way for AI models to access context beyond their training data—reading files, querying databases, calling APIs, and interacting with development tools.

Why MCP exists

Large language models are trained on static datasets and can't directly access real-time information or take actions in the world. Without additional infrastructure, an AI assistant can't:

• Read files from your local filesystem
• Query your company's internal documentation
• Execute code or run tests
• Interact with external APIs and services

MCP solves this by defining a protocol for AI systems to request and receive information from external sources, while maintaining security boundaries and user control.

How MCP works

MCP uses a client-server architecture:

1. MCP Servers expose capabilities (tools, resources, prompts) through a standardized JSON-RPC interface
2. MCP Clients (AI assistants, IDEs, development tools) connect to servers and invoke their capabilities
3. The protocol defines how clients discover available capabilities, request actions, and receive results

A developer might run MCP servers that provide access to:

• Local filesystem operations
• Git repository information
• Database queries
• Internal API documentation
• Code analysis tools

MCP in code review

MCP enables AI coding assistants to participate more effectively in code review workflows:

• Reading code context: AI can access the full codebase, not just the current file
• Running analysis tools: AI can invoke static analysis or testing tools and interpret results
• Understanding project structure: AI can query build systems, dependency graphs, and configuration
• Acting on feedback: When AI code review tools provide structured feedback (like a PR Report Card), MCP allows AI assistants to read and respond to that feedback

Security considerations

MCP includes mechanisms for:

• Capability discovery: Clients learn what servers can do before invoking actions
• User consent: Actions can require explicit user approval
• Sandboxing: Servers can limit what operations they expose

However, MCP servers have significant access to system resources. Running untrusted servers or granting excessive permissions creates security risk.

Adoption

MCP is supported by AI coding tools including Claude, Cursor, and various IDE extensions. The protocol is open source and designed for extension by the developer community.

See also: AI Code Review, PR Report Card