Home / MCP / Task Manager MCP Server

Task Manager MCP Server

Provides a Task Manager MCP server with planning, task tracking, UI, and review workflow for feature development.

typescript

21stars

View on GitHub

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
    "mcpServers": {
        "task_manager_mcp_http": {
            "url": "https://mcp.example.com/mcp"
        }
    }
}

You will run a Task Manager MCP Server that integrates with your AI editor to plan, track, and review code tasks with an unlimited context window. It includes a web UI to view tasks and progress, and it supports automatic and manual task orchestration to keep your development flow smooth and organized.

How to use

Set up the MCP server to work with your editor. You can run the server locally for testing or connect to a remote MCP endpoint. Use the editor’s MCP configuration to point to either the HTTP endpoint or the local stdio command, and then interact with the task planning, progress tracking, and reviews through the UI.

From your editor, configure the MCP server in a project or globally so the editor can start the server process and pass API keys and ports. When you request a feature, the server plans the implementation, creates tasks, and updates the UI with progress. If clarification is needed, the server pauses and presents a UI question to resolve context before continuing.

The server exposes basic UI endpoints to fetch features and tasks, create and update tasks, and check for active clarification questions. Use the UI to review task lists, adjust plans, review code changes, and see the progress of each feature as tasks are completed one by one.

How to install

Prerequisites: Node.js, npm, and Git installed on your machine.

1. Clone the project repository.

git clone https://github.com/jhawkins11/task-manager-mcp.git
cd task-manager-mcp

2. Install backend dependencies.

npm install

3. Build the backend and frontend, which also prepares the Svelte UI in dist/frontend-ui.

npm run build

4. Start the server locally for testing.

Option A — Run with Node directly (absolute path to the built server):

node /full/path/to/your/task-manager-mcp/dist/server.js

Option B — Use the npm start script (runs the MCP server, WebSocket server, and UI server):

npm start

The UI will be available at http://localhost:4999 by default unless you set a different UI_PORT in your environment.

Additional notes

Configuration is typically provided via environment variables when running the server directly or through your editor’s MCP settings. The server supports an OpenRouter-based workflow and can fall back to alternative Gemini models if needed. You may also configure the UI and WebSocket ports using environment variables.

Configuration details

{
  "mcpServers": {
    "task-manager-mcp": {
      "command": "node",
      "args": ["/full/path/to/your/task-manager-mcp/dist/server.js"],
      "env": {
        "OPENROUTER_API_KEY": "sk-or-v1-xxxxxxxxxxxxxxxxxxxx",
        // optional: you may specify preferred models
        // "OPENROUTER_MODEL": "google/gemini-2.5-flash-preview:thinking",
        // "FALLBACK_OPENROUTER_MODEL": "google/gemini-2.5-flash-preview:thinking",
        // optional: ports (default 4999)
        // "UI_PORT": "4999",
        // "WS_PORT": "4999"
        // "AUTO_REVIEW_ON_COMPLETION": "false"
      }
    }
  }
}

This configuration is used when running the server directly or via your editor. It passes the API key and optional model settings to the server at startup.

Available tools

plan_feature

Plan a new feature by generating a step-by-step development plan with recursive task breakdown for high-effort tasks.

mark_task_complete

Mark the current task as completed and move to the next pending task.

get_next_task

Retrieve the next task to work on after a completion or clarification.

review_changes

Review the code changes, potentially generate new tasks based on the review.

adjust_plan

Adjust the current plan after its creation using the adjust_plan tool.