The Dify MCP Client, developed by the 3dify Project, is a free Agent Strategy Plugin for the Dify platform. Built primarily with Python, this tool enables Dify to interact with external Model Context Protocol (MCP) servers, acting as an MCP client. It extends Dify's capabilities by converting external tools, resources, and prompts into Dify Tools accessible by Large Language Models. This integration allows Dify workflows to connect with a variety of external services and AI clients, enhancing automation and specialized functionality.

What does Dify MCP Client connect to?

The Dify MCP Client is specifically designed to connect with MCP (Model Context Protocol) servers, functioning as a critical Agent Strategy Plugin. This integration allows Dify to extend its operational scope by interacting with external MCP services. The client supports connections to standard input/output MCP servers, ensuring broad compatibility. For communication, Dify, when acting as an MCP client, supports both Streamable HTTP (which is the recommended option) and SSE (Server-Sent Events) transport protocols. This dual protocol support offers flexibility for diverse network environments.

Core features and tools

As an agent strategy plugin, Dify MCP Client converts lists of tools, resources, and prompts received from connected MCP servers directly into Dify Tools. This conversion makes them fully accessible for Large Language Model (LLM) utilization within the Dify platform. It exposes two primary callable tools within Dify: mcp_list_tools, which is used to discover the capabilities and available functions of the connected server, and mcp_call_tool, designed to execute specific functions or operations on that server. This enables dynamic interaction and utilization of external services directly from Dify's environment.

Who is it for?

The Dify MCP Client targets AI engineers, developers, and users who aim to integrate Dify workflows with external tools or expose Dify applications to other AI clients. It supports connecting Dify to specialized external capabilities and automating various tasks through interactions with MCP servers. Confirmed compatible AI clients that can interact with these MCP servers include Cursor, Claude Desktop, Cline, Windsurf, Cherry Studio, and other Dify instances. Skip this if your workflow does not involve external MCP server interactions or Dify integration with other AI clients.

How to get started

Access is granted through a standard OAuth authorization flow or via API keys. When connecting Dify to an external MCP server, the setup process involves providing the server URL and potentially custom HTTP headers for authentication, such as an "Authorization: Bearer <token>". Dify also supports dynamic, per-user MCP OAuth authorization, offering adaptable security configurations.

Use cases

• Integrating with various MCP servers for task automation.
• Utilizing the ReAct loop for dynamic agent interactions within Dify.
• Connecting Dify to external specialized capabilities, such as Notion for document search.

Dify MCP Client offers a crucial bridge for Dify users needing to extend their applications with external MCP server capabilities. It enables Dify to act as a versatile client, integrating diverse tools and services. The plugin supports various authentication methods and transport protocols, making it adaptable for different environments.