mcp-difyworkflow-server is an MCP (Model Context Protocol)-based server tool application, whose main purpose is to query and invoke workflows on the Dify platform. It allows users to run multiple custom Dify workflows on demand, thereby integrating Dify's workflow capabilities into MCP-enabled clients or environments.
Main capabilities include:
- List authorized workflows: view the list of available workflows under the current configuration through the list_workflows tool.
- Execute a specified workflow: execute a workflow through the execute_workflow tool, requiring the workflow name (workflow_name) and input content (input). Note that the input variable name of a Dify workflow should by default be defined as "message".
Usage:
- Configuration: you need to set the Dify platform's API server base URL, for example "base-url":"http://localhost/v1"。
- Command: you can specify the absolute path of the compiled binary file, or create a symbolic link, for example: sudo ln -s <gitWorkPath>/mcp-difyworkflow-server /usr/local/bin/mcp-difyworkflow-server, and then use "command":"mcp-difyworkflow-server".
- Environment variables: you need to configure DIFY_WORKFLOW_NAME and DIFY_API_KEYS, and the two must correspond one-to-one in order. The workflow name can be customized by the user in the Prompt, while the workflow API Key needs to be generated on the Dify platform for the corresponding workflow (for the specific generation method, please refer to the relevant Dify documentation).
Applicable scenarios:
- In MCP-enabled clients, invoke Dify workflows through natural language instructions, for example: "View the list of tools available to mcp-difyworkflow-server" or "Execute for me the dify workflow named workflow-translator, with the input message being 'This is a test message'".
- Need to integrate Dify workflows as tools into other AI applications or automation processes.
Limitations:
- Workflows need to be created in advance on the Dify platform and corresponding API Keys generated.
- The workflow input variable name must by default be "message".
- Other specific limitations (such as platform compatibility, performance metrics, etc.) are not provided for now.