Unified Operational Interface for DolphinScheduler (MCP Server)

A Model Context Protocol (MCP) implementation serving as an intermediary for programmatic interaction with Apache DolphinScheduler, enabling AI entities to control scheduling and execution functions using a consistent interface.

Core Concept

This component establishes a FastMCP server wrapper that exposes the underlying DolphinScheduler REST API as a collection of discrete, protocol-compliant operational functions consumable by autonomous agents. It functions as the essential communication conduit facilitating AI-driven management of distributed workflows.

Key Capabilities

• Complete exposure of all DolphinScheduler programmatic functionalities via the API.
• Adherence to the Model Context Protocol specification for tool invocation consistency.
• Simplified setup and parameterization using environment variables or command-line arguments.
• Built-in support for exhaustive tool documentation retrieval.

Deployment Instructions

Installation via Python package manager:

pip install dolphinscheduler-mcp

Configuration Parameters

Configuration is primarily managed via environment settings:

• DOLPHINSCHEDULER_API_URL: Endpoint address for the DolphinScheduler communication layer (Default: http://localhost:12345/dolphinscheduler).
• DOLPHINSCHEDULER_API_KEY: Security credential (token) required for authenticating requests against the DolphinScheduler service.
• DOLPHINSCHEDULER_MCP_HOST: Network interface address where the MCP service will listen (Default: 0.0.0.0).
• DOLPHINSCHEDULER_MCP_PORT: Network port assigned for the MCP service listener (Default: 8089).
• DOLPHINSCHEDULER_MCP_LOG_LEVEL: Verbosity setting for system logging (Default: INFO).

Operational Guidance

Execution via CLI

Launch the abstraction layer service using the command-line interface:

ds-mcp --host 0.0.0.0 --port 8089

Programmatic Server Initialization

Utilizing the internal Python interface:

from dolphinscheduler_mcp.server import run_server

# Initiate the service listener
run_server(host="0.0.0.0", port=8089)

Exposed Functional Toolsets

The DolphinScheduler MCP Server exposes functions categorized for interaction with the following DolphinScheduler domains:

• Project Structure Administration
• Process Blueprint Definition Control
• Executed Process Instance Oversight
• Individual Task Blueprint Management
• Workflow Scheduling Governance
• System Resource Inventory
• Data Connection Configuration
• Alerting Group Configuration
• Alerting Mechanism Plugin Administration
• Worker Cluster Group Management
• Multi-tenancy/Namespace Administration
• User Identity Management
• System Health and Metric Reporting

Illustrative Agent Interaction (Client Side)

from mcp_client import MCPClient

# Establish a connection channel to the listening service
client = MCPClient("http://localhost:8089/mcp")

# Query the repository for defined projects
response = await client.invoke_tool("get-project-list")

# Initiate the creation of a novel project entity
response = await client.invoke_tool(
    "create-project", 
    {"name": "My AI Project", "description": "Project created by AI"}
)

Governing Terms

Released under the terms of the Apache License, Version 2.0.