📌Overview

Purpose: The MCP Client serves as an agent strategy plugin that enables communication between Dify and MCP servers, facilitating effective tool utilization and resource management.

Overview: Dify's MCP Client transforms the agent's interaction with machine learning models and external resources by leveraging the ReAct framework. It allows for seamless connections to multiple MCP servers, enabling the execution of commands based on the agent's reasoning and observation processes.

Key Features:

Multi-server Support: Connects to multiple MCP servers simultaneously, enhancing flexibility and efficiency in executing tasks across different systems.
Dify Tools Integration: Converts Tool, Resource, and Prompt lists into accessible Dify Tools for the LLM, enabling an intuitive interface and direct tool invocation during the agent’s operations.
Customizable Configuration: Utilizes a JSON-based configuration for server connections, making deployment and updates straightforward.
Error Handling & Flexibility: Offers a streamlined installation process and error mitigation strategies, ensuring reliable plugin usage within the Dify ecosystem.

dify-mcp-client

MCP Client as Agent Strategy Plugin.

Important:
Dify is not MCP Server but MCP Host.

How it works

Each MCP client (ReAct Agent) node can connect to MCP servers.

Tool, Resource, and Prompt lists are converted into Dify Tools.
Your selected LLM can see their name, description, and argument type.
The LLM calls Tools based on the ReAct loop (Reason → Act → Observe).

Note:
Most of the code in this repository contains files similar to Dify Official Plugins / Agent Strategies found here:
https://github.com/langgenius/dify-official-plugins/tree/main/agent-strategies/cot_agent

What I did

Copied ReAct.py and renamed the file to mcpReAct.py.
Added config_json GUI input field by editing mcpReAct.yaml and class mcpReActParams().

Changes in `mcpReAct.py`

Added 12 new functions for MCP.
Implemented __init__() to initialize AsyncExitStack and event loop.
Modified _handle_invoke_action() with MCP-specific code.
Added MCP setup and cleanup in _invoke().

Important:
The ReAct while loop remains unchanged.

Update history

Add SSE MCP client (v0.0.2)
Support multi SSE servers (v0.0.3)
Update python module and simplify dependencies (v0.0.4)
- mcp(v1.1.2 → v1.6.0+)
- dify_plugin(0.0.1b72 → v0.1.0)

Caution and Limitation

Caution:
This plugin does not implement a human-in-the-loop mechanism by default, so only connect to reliable MCP servers.
To reduce risk, decrease max iterations (default: 3) to 1 and use this Agent node repeatedly in Chatflow.
Note agent memory resets at the end of the Workflow. Use Conversation Variables to save history and pass it to QUERY.
Also, include instructions such as "ask for user's permission when calling tools" in the INSTRUCTION.

How to use this plugin

Install the plugin from GitHub

Enter the GitHub repository name:

https://github.com/3dify-project/dify-mcp-client/

In Dify: PLUGINS > + Install plugin > INSTALL FROM > GitHub

Install the plugin from `.difypkg` file

Go to Releases: https://github.com/3dify-project/dify-mcp-client/releases
Select a suitable version of .difypkg.
In Dify: PLUGINS > + Install plugin > INSTALL FROM > Local Package File

Handling plugin installation errors

Issue: If you encounter the error:
plugin verification has been enabled, and the plugin you want to install has a bad signature

Solution:

Open /docker/.env and change:

FORCE_VERIFYING_SIGNATURE=false

Restart Dify services:

cd docker
docker compose down
docker compose up -d

This disables signature verification, allowing installation of unverified plugins.

Tip:
Marketplace approval may be required. If stars⭐ reach 100, author may consider making a PR for marketplace inclusion.

Where does this plugin show up?

Takes a few minutes to install.
Once installed, use it in any workflow as an Agent node.
Select the "mcpReAct" strategy (otherwise, no MCP functionality).

Configuration

MCP Agent Plugin node requires a config_json to specify commands or URL(s) to connect MCP servers:

{
    "mcpServers": {
        "name_of_server1": {
            "url": "http://host.docker.internal:8080/sse"
        },
        "name_of_server2": {
            "url": "http://host.docker.internal:8008/sse"
        }
    }
}

Warning:

Each server's port number should be unique (e.g., 8080, 8008, etc.).

To use a stdio MCP server, consider:

Converting it to an SSE MCP server

Deploying with source code (not via .difypkg or GitHub repository install)

Pre-installing Node.js inside the dify-plugin docker container (see issue #10 in repository)

Chatflow example

The Tools field should not be blank. Select Dify tools like "current time".
A sample ChatFlow .yml for testing is available in the repository's test/chatflow folder.
After downloading the DSL(yml) file, import it in Dify and test MCP using "Everything MCP server".
See: https://github.com/modelcontextprotocol/servers/tree/main/src/everything

How to convert `stdio` MCP server into SSE MCP server

Option 1: Edit MCP server's code

If using fastMCP server, change:

if __name__ == "__main__":
-    mcp.run(transport="stdio")
+    mcp.run(transport="sse")

Option 2: via mcp-proxy

\mcp-proxy> uv venv -p 3.12
.venv\Scripts\activate
uv tool install mcp-proxy

Check Node.js and npx (.cmd) path:
(Mac/Linux)

which npx

(Windows)

where npx

Example results:

C:\Program Files\nodejs\npx
C:\Program Files\nodejs\npx.cmd
C:\Users\USER_NAME\AppData\Roaming\npm\npx
C:\Users\USER_NAME\AppData\Roaming\npm\npx.cmd

If claude_desktop_config.json is:

{
  "mcpServers": {
    "SERVER_NAME": {
       "command": CMD_NAME_OR_PATH,
       "args": {VALUE1, VALUE2}
    }
  }
}

Wake up stdio MCP server by:

mcp-proxy --sse-port=8080 --pass-environment -- CMD_NAME_OR_PATH --arg1 VALUE1 --arg2 VALUE2 ...

(On Windows, use npx.cmd instead of npx.)

Example to convert stdio "everything MCP server" to SSE via mcp-proxy:

mcp-proxy --sse-port=8080 --pass-environment -- C:\Program Files\nodejs\npx.cmd --arg1 -y --arg2 @modelcontextprotocol/server-everything

Another example using sample Chatflow for v0.0.3:

pip install mcp-simple-arxiv
mcp-proxy --sse-port=8008 --pass-environment -- C:\Users\USER_NAME\AppData\Local\Programs\Python\Python310\python.exe -m -mcp_simple_arxiv

Example mcp-proxy setup log:

(mcp_proxy) C:\User\USER_NAME\mcp-proxy>mcp-proxy --sse-port=8080 --pass-environment -- C:\Program Files\nodejs\npx.cmd --arg1 -y --arg2 @modelcontextprotocol/server-everything
DEBUG:root:Starting stdio client and SSE server
DEBUG:asyncio:Using proactor: IocpProactor
DEBUG:mcp.server.lowlevel.server:Initializing server 'example-servers/everything'
DEBUG:mcp.server.sse:SseServerTransport initialized with endpoint: /messages/
INFO:     Started server process [53104]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://127.0.0.1:8080 (Press CTRL+C to quit)

How to develop and deploy plugin

Official plugin development guide

https://github.com/3dify-project/dify-mcp-client/blob/main/GUIDE.md

Dify plugin SDK daemon

If your OS is Windows and CPU is Intel or AMD, download dify-plugin-windows-amd64.exe (v0.0.7) from:
https://github.com/langgenius/dify-plugin-daemon/releases
Steps:

Rename it to dify.exe for convenience.
Create folder C\User\user\.local\bin (Windows) and register it in system PATH.
Copy dify.exe to under dify-mcp-client/.

Tip:
Helpful guide: https://docs.dify.ai/plugins/quick-start/develop-plugins/initialize-development-tools

Reference

https://docs.dify.ai/plugins/quick-start/develop-plugins/initialize-development-tools

Note:
You can skip this stage if you pull or download code of this repo using:

dify plugin init

Initial settings will be prepared automatically.

Change directory

cd dify-mcp-client

Install Python modules

Python 3.12+ is compatible. Using virtual environments (venv) and uv tool is recommended but not necessary.

uv venv -p 3.12
.venv\Scripts\activate

Install python modules required for plugin development:

uv pip install -r requirements.txt

Duplicate `env.example` and rename to `.env`

I changed REMOTE_INSTALL_HOST from debug.dify.ai to localhost for Docker Compose environment.

Activate Dify plugin

python -m main

Use Ctrl+C to stop the plugin.

Tip:
REMOTE_INSTALL_KEY in .env often changes. If you get errors like handshake failed, invalid key, renew it.

Package into `.difypkg`

Default root is ./dify-mcp-client:

dify plugin package ./ROOT_OF_YOUR_PROJECT

Useful GitHub repositories for developers

Dify Plugin SDKs

https://github.com/langgenius/dify-plugin-sdks

MCP Python SDK

https://github.com/modelcontextprotocol/python-sdk

Tip:
MCP client example:
https://github.com/modelcontextprotocol/python-sdk/blob/main/examples/clients/simple-chatbot/mcp_simple_chatbot/main.py

Note:
Dify plugin requirements.txt automatically installs python modules, including the latest mcp package, so no need to download MCP SDK separately.

dify-mcp-client

📌Overview

dify-mcp-client

How it works

What I did

Changes in mcpReAct.py

Update history

Caution and Limitation

How to use this plugin

Install the plugin from GitHub

Install the plugin from .difypkg file

Handling plugin installation errors

Where does this plugin show up?

Configuration

Chatflow example

How to convert stdio MCP server into SSE MCP server

Option 1: Edit MCP server's code

Option 2: via mcp-proxy

How to develop and deploy plugin

Official plugin development guide

Dify plugin SDK daemon

Reference

Change directory

Install Python modules

Duplicate env.example and rename to .env

Activate Dify plugin

Package into .difypkg

Useful GitHub repositories for developers

Dify Plugin SDKs

MCP Python SDK

Changes in `mcpReAct.py`

Install the plugin from `.difypkg` file

How to convert `stdio` MCP server into SSE MCP server

Duplicate `env.example` and rename to `.env`

Package into `.difypkg`