
Task Orchestration
Task Orchestration
README
Task Orchestrator
A Model Context Protocol (MCP) server for task orchestration and management. This tool helps break down goals into manageable tasks and track their progress.
How to use
Ideally, the LLM should be able to understand when this MCP tool should be used. But as a sample prompt, something like this can possibly work
"Create a new development goal for me. The goal is to 'Implement user authentication' and it's for the 'my-web-app' repository."
LEMME KNOW of any issues you face in the 'discussions' tab.
Features
- Create and manage goals
- Break down goals into hierarchical tasks
- Track task completion status
- Support for subtasks and dependency management between parent task and subtasks
- Persistent storage using LokiDB
Roadmap
- Complex task/goal inter-dependency orchestration
- Goal deletion
- Completion dispositions
- UI for visualization of progress
API Reference
Task ID Naming Convention
Task IDs use a dot-notation (e.g., "1", "1.1", "1.1.1") where each segment represents a level in the hierarchy.
- For each new goal, top-level task IDs start with "1" and increment sequentially (e.g., "1", "2", "3").
- Subtasks have IDs formed by appending a new segment to their parent's ID (e.g., "1.1" is a subtask of "1").
- The combination of
goalId
andtaskId
is guaranteed to be unique.
Tools
The server provides the following tools (based on build/index.js
):
-
create_goal
- Create a new goal
- Parameters:
{ description: string; // The goal description repoName: string; // The repository name associated with this goal }
- Sample Input:
{ "description": "Implement user authentication", "repoName": "example/auth-service" }
- Returns:
{ goalId: number }
-
add_tasks
- Add multiple tasks to a goal. Tasks can be provided in a hierarchical structure. For tasks that are children of existing tasks, use the
parentId
field. The operation is transactional: either all tasks in the batch succeed, or the entire operation fails. - Parameters:
{ goalId: number; // ID of the goal to add tasks to (number) tasks: Array<{ title: string; // Title of the task (string) description: string; // Detailed description of the task (string) parentId?: string | null; // Optional parent task ID for tasks that are children of *existing* tasks. Do not use for new subtasks defined hierarchically within this batch. subtasks?: Array<any>; // An array of nested subtask objects to be created under this task. }>; }
- Sample Input:
{ "goalId": 1, "tasks": [ { "title": "Design database schema", "description": "Define tables for users, roles, and permissions", "subtasks": [ { "title": "Create ERD", "description": "Draw entity-relationship diagram" } ] }, { "title": "Implement user registration", "description": "Create API endpoint for new user signup", "parentId": "1" } ] }
- Returns:
HierarchicalTaskResponse[]
.HierarchicalTaskResponse
objects are simplified and do not includecreatedAt
,updatedAt
, orparentId
.
- Add multiple tasks to a goal. Tasks can be provided in a hierarchical structure. For tasks that are children of existing tasks, use the
-
remove_tasks
- Soft-delete multiple tasks from a goal. Tasks are marked as deleted but remain in the system. By default, a parent task with subtasks cannot be soft-deleted without explicitly deleting its children. Soft-deleted tasks are excluded by default from
get_tasks
results unlessincludeDeletedTasks
is set to true. - Parameters:
{ goalId: number; // ID of the goal to remove tasks from taskIds: string[]; // IDs of the tasks to remove (array of strings). Task IDs use dot-notation (e.g., "1", "1.1"). deleteChildren?: boolean; // Whether to delete child tasks along with the parent (boolean). Defaults to false. If false, attempting to delete a parent task with existing subtasks will throw an error. }
- Sample Input (without deleting children):
{ "goalId": 1, "taskIds": ["2", "3"] }
- Sample Input (with deleting children):
{ "goalId": 1, "taskIds": ["1"], "deleteChildren": true }
- Returns:
{ removedTasks: TaskResponse[], completedParents: TaskResponse[] }
.TaskResponse
objects are simplified and do not includecreatedAt
,updatedAt
, orparentId
.
- Soft-delete multiple tasks from a goal. Tasks are marked as deleted but remain in the system. By default, a parent task with subtasks cannot be soft-deleted without explicitly deleting its children. Soft-deleted tasks are excluded by default from
-
get_tasks
- Get tasks for a goal. Task IDs use a dot-notation (e.g., "1", "1.1", "1.1.1"). When
includeSubtasks
is specified, responses will return hierarchical task objects. Otherwise, simplified task objects withoutcreatedAt
,updatedAt
, orparentId
will be returned. - Parameters:
{ goalId: number; // ID of the goal to get tasks for (number) taskIds?: string[]; // Optional: IDs of tasks to fetch (array of strings). If null or empty, all tasks for the goal will be fetched. includeSubtasks?: "none" | "first-level" | "recursive"; // Level of subtasks to include: "none" (only top-level tasks), "first-level" (top-level tasks and their direct children), or "recursive" (all nested subtasks). Defaults to "none". includeDeletedTasks?: boolean; // Whether to include soft-deleted tasks in the results (boolean). Defaults to false. }
- Sample Input:
{ "goalId": 1, "includeSubtasks": "recursive", "includeDeletedTasks": true }
- Returns:
TaskResponse[]
.TaskResponse
objects are simplified and do not includecreatedAt
,updatedAt
, orparentId
.
- Get tasks for a goal. Task IDs use a dot-notation (e.g., "1", "1.1", "1.1.1"). When
-
complete_task_status
- Mark tasks as complete. By default, a parent task cannot be marked complete if it has incomplete child tasks.
- Parameters:
{ goalId: number; // ID of the goal containing the tasks taskIds: string[]; // IDs of the tasks to update (array of strings). Task IDs use dot-notation (e.g., "1", "1.1"). completeChildren?: boolean; // Whether to complete all child tasks recursively (boolean). Defaults to false. If false, a task can only be completed if all its subtasks are already complete. }
- Sample Input (without completing children):
{ "goalId": 1, "taskIds": ["1", "2"] }
- Sample Input (with completing children):
{ "goalId": 1, "taskIds": ["1"], "completeChildren": true }
- Returns:
TaskResponse[]
.TaskResponse
objects are simplified and do not includecreatedAt
,updatedAt
, orparentId
.
Usage Examples
Creating a Goal and Tasks
// Create a new goal. Its top-level tasks will start with ID "1".
const goal = await callTool('create_goal', {
description: 'Implement user authentication',
repoName: 'user/repo'
});
// Add a top-level task
const task1 = await callTool('add_tasks', {
goalId: goal.goalId,
tasks: [
{
title: 'Set up authentication middleware',
description: 'Implement JWT-based authentication'
}
]
});
// task1.addedTasks[0].id will be "1"
// Add a subtask to the previously created task "1"
const task2 = await callTool('add_tasks', {
goalId: goal.goalId,
tasks: [
{
title: 'Create login endpoint',
description: 'Implement POST /auth/login',
parentId: "1" // ParentId must refer to an *already existing* task ID
}
]
});
// task2.addedTasks[0].id will be "1.1"
Managing Task Status
// Mark a parent task as complete, which will also complete its children
await callTool('complete_task_status', {
goalId: 1,
taskIds: ["1"],
completeChildren: true
});
// Get all tasks including subtasks recursively
const allTasks = await callTool('get_tasks', {
goalId: 1,
includeSubtasks: "recursive"
});
Removing Tasks
// Attempt to remove a parent task without deleting children (will fail if it has subtasks)
try {
await callTool('remove_tasks', {
goalId: 1,
taskIds: ["1"]
});
} catch (error) {
console.error(error.message); // Expected to throw an error if subtasks exist
}
// Remove a parent task and its children
await callTool('remove_tasks', {
goalId: 1,
taskIds: ["1"],
deleteChildren: true
});
Development
Prerequisites
- Node.js 18+
- pnpm
Setup
-
Install dependencies:
pnpm install
-
Build the project:
pnpm build
-
Run tests:
pnpm test
Project Structure
src/
- Source codeindex.ts
- Main server implementationstorage.ts
- Data persistence layertypes.ts
- TypeScript type definitionsprompts.ts
- AI prompt templates__tests__/
- Test files
License
MIT
推荐服务器

Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。