Dify Workflow MCP Server

Overview

The mcp-difyworkflow-server is a Model Context Protocol (MCP) server that bridges AI assistants with the Dify platform, enabling on‑demand execution of custom workflows directly from conversational prompts. By exposing two core tools— and —the server allows developers to discover available Dify workflows and trigger them with user‑supplied inputs without leaving the AI chat environment. This eliminates the need for separate API calls or manual orchestration, streamlining the integration of complex business logic into AI‑driven applications.

For developers building conversational agents, this server solves a key pain point: the cumbersome coupling between an AI assistant and external workflow engines. Dify workflows often encapsulate multi‑step processes such as data transformation, external API calls, or business rule enforcement. The MCP server translates a simple prompt into an authenticated request to the Dify API, automatically handling authentication via per‑workflow API keys and routing the input payload. This tight coupling means that a user can, for example, ask an assistant to “translate this text” and the underlying Dify workflow will perform language translation, return the result, and pass it back to the assistant—all transparently.

Key capabilities of the server include:

• Dynamic workflow discovery: returns a catalog of workflows the server has permission to invoke, making it easy for users to see what actions are available.
• Parameter mapping: The server expects the workflow’s input variable to be named , ensuring consistent payload structure across workflows.
• Parallel execution support: Multiple workflows can be defined and invoked concurrently, each with its own API key, allowing a single assistant to orchestrate complex pipelines.
• Secure configuration: API keys are supplied via environment variables, keeping secrets out of the codebase and enabling fine‑grained access control per workflow.

Typical use cases include:

• Customer support automation: An assistant can forward user queries to a Dify workflow that routes the message through sentiment analysis, knowledge base lookup, and ticket creation.
• Content generation pipelines: A workflow that takes a prompt, runs it through multiple generative models, and aggregates the outputs can be triggered from within a chat.
• Data enrichment: User inputs are enriched with external data sources (e.g., stock prices, weather) via a Dify workflow before being returned.

Integration into an AI workflow is straightforward: the MCP server registers its tools with the client’s tool registry, and prompts can reference or . The assistant’s natural language understanding layer maps user intent to the appropriate tool, passes the required arguments, and renders the workflow’s output as part of the conversation. This seamless flow turns a static AI model into an orchestrated service that can leverage sophisticated backend logic with minimal friction.