MCP Orchestrator Server

What is MCP Orchestrator Server?

A central coordination server that orchestrates tasks across multiple specialized MCP services, such as Trello and GitHub. It is designed for AI agents and developers who need to automate cross-service workflows via a unified JSON-RPC interface.

How to use MCP Orchestrator Server?

Install dependencies with npm install, then ensure the Trello and GitHub MCP servers are running (e.g., via Docker on ports 3001 and 3002). Start the orchestrator with npm run dev (development) or npm run build && npm start (production). Invoke its tools (planDevelopmentFeature, executeMultiServiceTask) via standard MCP tool calls.

Key features of MCP Orchestrator Server

• Supports planDevelopmentFeature for end-to-end feature rollout.
• Supports executeMultiServiceTask for custom multi-service workflows.
• Integrates with Trello and GitHub MCP servers out of the box.
• Provides robust error handling (timeouts, retries, partial responses).
• Logs detailed results with per-operation success/failure info.
• Designed to be extended with additional MCP services via code.

Use cases of MCP Orchestrator Server

• Automatically create a Trello card and a GitHub feature branch from a single instruction.
• Execute a sequence of actions across different services (e.g., create a Trello list then a GitHub branch).
• Enable AI agents to translate natural language requests into structured multi-step plans.
• Coordinate multiple MCP services in a scalable, centralized manner.

FAQ from MCP Orchestrator Server

What is the difference between planDevelopmentFeature and executeMultiServiceTask?

planDevelopmentFeature is a high-level tool that automates the common pattern of creating a Trello backlog, card, and GitHub branch for a new feature. executeMultiServiceTask gives you full control to define an arbitrary array of tasks across different services.

Which MCP services are currently supported?

The orchestrator natively supports Trello MCP and GitHub MCP servers. The README notes that GitHub MCP is “not implemented yet” in the architecture diagram, but the examples and available tools use both services.

What are the runtime requirements to run the orchestrator?

You need Node.js, npm, and both the Trello and GitHub MCP servers running (e.g., via Docker) before starting the orchestrator. No other external dependencies are mentioned.

How does the orchestrator handle errors?

It includes automatic timeouts and retries for network failures, continues with available services if one is down, reports validation errors with the failing parameter, and returns partial responses indicating which operations succeeded.

Where does data live and how is authentication handled?

The README does not specify where data is stored or authentication details. It only configures base URLs for other MCP services (e.g., http://localhost:3001), implying local or network-accessible endpoints are used without additional auth layers described.