Model Context Protocol (MCP)
Model Context Protocol (MCP) is an open protocol specification introduced by Anthropic in 2024 that standardizes the interface between AI models and the external tools, APIs, and data sources they can use; An MCP server is a lightweight process that exposes a manifest of available tools with JSON Schema parameter definitions; MCP reduces the bespoke integration work required to connect LLMs to external systems
Model Context Protocol (MCP) is an open protocol specification introduced by Anthropic in 2024 that standardizes the interface between AI models and the external tools, APIs, and data sources they can use. MCP defines a client-server architecture where MCP servers expose capabilities (tools, resources, prompts) and MCP clients (AI applications) discover and invoke them through a consistent message format.
How it works
An MCP server is a lightweight process that exposes a manifest of available tools with JSON Schema parameter definitions. The MCP client, typically an AI agent host, queries the server for available tools and passes them to the model. When the model requests a tool call, the client routes it to the appropriate MCP server, receives the result, and returns it to the model. Communication uses JSON-RPC over stdio or HTTP with SSE.
Key facts
- Released: November 2024 by Anthropic as an open standard; adopted by OpenAI and other providers in 2025.
- Transports: stdio (for local tool servers) and HTTP/SSE (for remote servers) are the two supported transports.
- Resource types: MCP servers expose tools (callable functions), resources (readable data), and prompts (templated instructions).
- Ecosystem: Hundreds of MCP servers exist for services like GitHub, Slack, databases, file systems, and web browsers.
For builders
MCP reduces the bespoke integration work required to connect LLMs to external systems. Rather than writing custom function schemas for every tool, builders can deploy or reuse existing MCP servers. Claude Code, Cursor, and other AI development tools use MCP to give models access to file systems, terminals, and APIs within a controlled, auditable protocol.
Sources
- Anthropic. Model Context Protocol specification. modelcontextprotocol.io
- Yao, S., et al. (2022). ReAct: Synergizing Reasoning and Acting in Language Models. arXiv:2210.03629. arxiv.org
- Madaan, A., et al. (2023). Self-Refine: Iterative Refinement with Self-Feedback. arXiv:2303.17651. arxiv.org
- Anthropic. Model Context Protocol specification. modelcontextprotocol.io
- Anthropic. Tool use with Claude. docs.anthropic.com
