Ifly Workflow MCP Server

What is Ifly Workflow MCP Server about?

Provides a simple implementation of an MCP server that connects MCP tools with iFlytek workflow services, allowing developers to invoke complex AI‑driven workflows via the Model Context Protocol.

How to use Ifly Workflow MCP Server?

1. Create a config.yaml that lists workflow IDs, names, descriptions and api_key (format: API Key:API Secret).
2. Add the server definition to your MCP client configuration (claude_desktop_config.json or mcp.json), pointing to the repository and specifying the CONFIG_PATH environment variable.
3. Run the server through the configured command; MCP clients can now send requests to the defined workflows.

Key features of Ifly Workflow MCP Server

• Supports 14 node types (basic, tool, logic, transformation) with complex variable I/O.
• Execution modes: sequential, parallel, loop, and nested sub‑workflows.
• Hook mechanism enables streaming output for real‑time responses.
• Multiple development paradigms: single‑turn linear, branching, looped, and multi‑turn interactions.
• Multi‑model support via a Model‑of‑Models architecture, allowing model selection at critical stages.

Use cases of Ifly Workflow MCP Server

• Automating business processes that require AI decision making and data transformation.
• Building conversational agents that remember context across multiple turns.
• Implementing parallel data enrichment pipelines where several tools run simultaneously.
• Creating reusable sub‑workflows for common tasks such as validation, formatting, or external API calls.

FAQ from the Ifly Workflow MCP Server

Q: What authentication is needed? A: Provide api_key (API Key and API Secret) in the config.yaml for each workflow.

Q: How do I publish a workflow as an API? A: Debug the workflow, start a conversation to verify it works, then click Publish as API → Configure and bind an application to obtain the workflow ID and credentials.

Q: Can I run multiple workflows concurrently? A: Yes, the server supports parallel execution and nested sub‑workflows, enabling concurrent processing.

Q: Where should I place the configuration file? A: Set the CONFIG_PATH environment variable to the absolute path of your config.yaml when launching the server.

Q: Is streaming output available? A: Yes, the hook mechanism streams results back to the client as they are generated.

iFlytek Workflow MCP Server

The Model Context Protocol (MCP) is an open protocol designed for effortless integration between LLM applications and external data sources or tools, offering a standardized framework to seamlessly provide LLMs with the context they require.

This a simple implementation of an MCP server using iFlytek. It enables calling iFlytek workflows through MCP tools.

Features

Functional Overview

This system is built on the iFlytek MCP server and enables intelligent workflow scheduling, making it suitable for various business scenarios.

• Workflow Structure: Composed of multiple nodes, supporting 14 types of nodes (including basic, tool, logic, and transformation types).
• Core Components: By default, the workflow includes a Start Node (user input) and an End Node (output result).
• Execution Mode: Once triggered, the workflow executes automatically according to predefined sequences and rules, requiring no manual intervention.

Core Capabilities

Robust Node Support

• 14 types of workflow nodes to meet diverse business requirements.
• Supports complex variable I/O, enabling flexible data transmission.

Advanced Orchestration Modes

• Sequential Execution: Tasks execute one after another in order.
• Parallel Execution: Multiple tasks run simultaneously to enhance efficiency.
• Loop Execution: Supports iterative loops for handling repetitive tasks.
• Nested Execution: Allows embedding sub-workflows within workflows, improving reusability.
• Utilizes the Hook Mechanism to enable streaming output, ensuring real-time processing.

Multiple Development Paradigms

• Single-turn, single-branch: Linear execution of simple tasks.
• Single-turn, multi-branch: Supports branching logic to handle complex processes.
• Single-turn loop: Manages looped tasks to enhance automation.
• Multi-turn interaction: Supports context memory for dynamic conversations.

Capability Expansion

• Multi-Model Support: Based on the Model of Models (MoM) hybrid application architecture, providing multiple model choices at critical workflow stages. This allows for flexible model combinations, improving task adaptability.

Usage with MCP client

Prepare config.yaml

Before using the mcp server, you should prepare a config.yaml to save your workflow info. The example config like this:

- flow_id: 'flow id'              # required
  name: 'flow name'               # optional, if not set, obtain the name from the cloud.
  description: 'flow description' # optional, if not set, obtain the description from the cloud.
  api_key: 'API Key:API Secret'   # required

Get workflow authentication information

1. Create a bot

2. Publish a workflow

• Step 1. Debug the workflow you just created.
• Step 2. Engage in a conversation with your workflow and ensure the conversation is successful.
• Step 3. You can now click the publish button.
• Step 4. Select "Publish as API" and click the "Configure" button.
• Step 5. Select the application you need to bind and bind it. Now you can retrieve the corresponding workflow ID and authentication information. Enjoy!

Note: If you find that you are unable to select an app, you can go to https://www.xfyun.cn to apply.

Manual Installation

To add a persistent client, add the following to your claude_desktop_config.json or mcp.json file:

{
    "mcpServers": {
        "ifly-workflow-mcp-server": {
            "command": "uvx",
            "args": [
                "--from",
                "git+https://github.com/iflytek/ifly-workflow-mcp-server",
                "ifly_workflow_mcp_server"
            ],
            "env": {
                "CONFIG_PATH": "$CONFIG_PATH"
            }
        }
    }
}

Example config:

{
    "mcpServers": {
        "ifly-workflow-mcp-server": {
            "command": "uvx",
            "args": [
                "--from",
                "git+https://github.com/iflytek/ifly-workflow-mcp-server",
                "ifly_workflow_mcp_server"
            ],
            "env": {
                "CONFIG_PATH": "/Users/hygao1024/Projects/config.yaml"
            }
        }
    }
}