Model Context Protocol (MCP) is an open, standardized protocol that enables seamless integration between large language model (LLM) applications and external data sources, tools, and workflows. MCP provides a robust, extensible framework for sharing context, exposing capabilities, and building composable AI-powered systems.
Whether you're building an AI IDE, enhancing a chat interface, or orchestrating complex agentic workflows, MCP is the universal connector for context-aware AI
- Standardized Context Sharing: Share structured context (user, system, task, document) with LLMs in a machine- and human-readable format.
- Tool and Resource Integration: Expose external tools, APIs, and data sources to LLMs in a safe, controlled way.
- Composable Workflows: Build modular, interoperable AI workflows using a common protocol.
- Model-Agnostic: Works with any LLM or agent framework, supporting a wide range of programming languages and platforms.
- Security & Consent: Built-in principles for user consent, data privacy, and safe tool execution.
- Extensible: Easily add new features, context types, and integrations as your needs evolve.
MCP uses a client-server architecture based on JSON-RPC 2.0 messages:
- Host: The LLM application (e.g., IDE, chat app) that initiates connections.
- Client: The connector within the host that communicates using MCP.
- Server: The service providing context, tools, and capabilities to the client.
MCP is inspired by the Language Server Protocol (LSP), but is designed for the broader AI ecosystem.
- Contextualize LLMs: Provide rich, structured context to models for more accurate, relevant, and safe responses.
- Expose Tools: Allow LLMs to call external functions, APIs, and workflows securely.
- Integrate Data: Connect LLMs to databases, filesystems, and real-time data streams.
- Enable Agentic Workflows: Orchestrate multi-step, multi-agent processes with shared context and tool access.
- AI IDEs: Enhance code editors with context-aware completions, refactoring, and tool integration.
- Chatbots & Assistants: Build smarter, safer conversational agents that can access tools and data.
- Enterprise Automation: Standardize how AI systems interact with business tools and workflows.
- Research & Education: Provide reproducible, explainable AI interactions for learning and experimentation.
MCP is designed with security and trust at its core:
- User Consent: Users must explicitly approve data sharing and tool execution.
- Data Privacy: Sensitive data is protected and never shared without permission.
- Tool Safety: All tool calls are controlled and auditable.
- Transparent Workflows: All context and actions are visible and explainable.
- Read the MCP Specification
- Explore SDKs: Python, TypeScript, Go, Java, Kotlin, C#, Rust, Swift, Ruby
- Try Example Servers: Reference Servers
- Join the Community: modelcontextprotocol.io
Model Context Protocol is open source under the MIT License. See LICENSE for details.
"MCP: The universal connector for context-aware AI."