MCP (Model Context Protocol)
Open protocol enabling AI models to interact with external tools and APIs in a standardized way.
Open protocol enabling AI models to interact with external tools and APIs in a standardized way.
Définition détaillée
The Model Context Protocol (MCP) is a standardized protocol developed by Anthropic that allows AI models (LLMs) to discover and interact with external tools, databases, and APIs in a structured manner. MCP defines a common format for tool descriptions (name, parameters, input/output schema) and an invocation mechanism. It simplifies AI agent integration with existing information systems: CRM, ERP, SQL databases, REST APIs, cloud services.
Cas d'usage
CRM-connected AI assistant: via MCP, the assistant can view customer records, update opportunities, create tasks, and send emails directly from the conversational interface.
Termes associés
En savoir plus
Questions fréquentes
Is MCP compatible with all LLMs?
MCP was designed by Anthropic for Claude, but open-source implementations enable its use with other models (OpenAI, Llama, Mistral).
What's the difference between MCP and native function calling?
MCP standardizes tool discovery and invocation across different models and providers, whereas function calling is specific to each provider's API.