@kazuph/mcp-taskmanager
MCP 任务管理器
用于任务管理的模型上下文协议服务器。这允许 Claude Desktop(或任何 MCP 客户端)在基于队列的系统中管理和执行任务。
快速入门(针对用户)
先决条件
Node.js 18+(通过
brew install node安装)Claude Desktop(从https://claude.ai/desktop安装)
配置
打开您的 Claude Desktop 配置文件:
~/Library/Application Support/Claude/claude_desktop_config.json
您可以通过 Claude 桌面菜单找到它:
打开 Claude 桌面
在 Mac 菜单栏上单击 Claude
点击“设置”
点击“开发者”
将以下内容添加到您的配置中:
{
"tools": {
"taskmanager": {
"command": "npx",
"args": ["-y", "@kazuph/mcp-taskmanager"]
}
}
}Related MCP server: Todoist MCP Server
对于开发人员
先决条件
Node.js 18+(通过
brew install node安装)Claude Desktop(从https://claude.ai/desktop安装)
tsx(通过
npm install -g tsx)
安装
git clone https://github.com/kazuph/mcp-taskmanager.git
cd mcp-taskmanager
npm install
npm run build开发配置
确保 Claude Desktop 已安装并正在运行。
如果还没有安装 tsx,请全局安装:
npm install -g tsx
# or
pnpm add -g tsx修改位于以下位置的 Claude Desktop 配置:
~/Library/Application Support/Claude/claude_desktop_config.json
将以下内容添加到您的 MCP 客户端的配置中:
{
"tools": {
"taskmanager": {
"args": ["tsx", "/path/to/mcp-taskmanager/index.ts"]
}
}
}可用操作
TaskManager 支持两个主要操作阶段:
规划阶段
接受用户的任务列表(字符串数组)
将任务以队列形式内部存储
返回执行计划(任务概述、任务ID、当前队列状态)
执行阶段
当请求时从队列返回下一个任务
提供任务完成的反馈机制
从队列中移除已完成的任务
准备执行下一个任务
参数
action:“计划” | “执行” | “完成”tasks:任务字符串数组(“计划”操作所需)taskId:任务标识符(“完成”操作所需)getNext:布尔标志,用于请求下一个任务(用于“执行”操作)
示例用法
// Planning phase
{
action: "plan",
tasks: ["Task 1", "Task 2", "Task 3"]
}
// Execution phase
{
action: "execute",
getNext: true
}
// Complete task
{
action: "complete",
taskId: "task-123"
}Available Tools
10 toolsadd_tasks_to_requestC
Add new tasks to an existing request. This allows extending a request with additional tasks.
A progress table will be displayed showing all tasks including the newly added ones.
Inputs: requestId (required), tasks (required).
approve_request_completionA
After all tasks are done and approved, this tool finalizes the entire request. The user must call this to confirm that the request is fully completed.
A progress table showing the final status of all tasks will be displayed before requesting final approval.
If not approved, the user can add new tasks using 'request_planning' and continue the process.
Inputs: requestId (required).
approve_task_completionA
Once the assistant has marked a task as done using 'mark_task_done', the user must call this tool to approve that the task is genuinely completed. Only after this approval can you proceed to 'get_next_task' to move on.
A progress table will be displayed before requesting approval, showing the current status of all tasks.
If the user does not approve, do not call 'get_next_task'. Instead, the user may request changes, or even re-plan tasks by using 'request_planning' again.
Inputs: requestId (required), taskId (required).
delete_taskB
Delete a specific task from a request. Only uncompleted tasks can be deleted.
A progress table will be displayed showing the remaining tasks after deletion.
Inputs: requestId (required), taskId (required).
get_next_taskA
Given a 'requestId', return the next pending task (not done yet). If all tasks are completed, it will indicate that no more tasks are left and that you must wait for the request completion approval.
A progress table showing the current status of all tasks will be displayed with each response.
If the same task is returned again or if no new task is provided after a task was marked as done but not yet approved, you MUST NOT proceed. In such a scenario, you must prompt the user for approval via 'approve_task_completion' before calling 'get_next_task' again. Do not skip the user's approval step. In other words:
After calling 'mark_task_done', do not call 'get_next_task' again until 'approve_task_completion' is called by the user.
If 'get_next_task' returns 'all_tasks_done', it means all tasks have been completed. At this point, you must not start a new request or do anything else until the user decides to 'approve_request_completion' or possibly add more tasks via 'request_planning'.
Inputs: requestId (required).
list_requestsB
List all requests with their basic information and summary of tasks. This provides a quick overview of all requests in the system.
Inputs: no named parameters.
mark_task_doneA
Mark a given task as done after you've completed it. Provide 'requestId' and 'taskId', and optionally 'completedDetails'.
After marking a task as done, a progress table will be displayed showing the updated status of all tasks.
After this, DO NOT proceed to 'get_next_task' again until the user has explicitly approved this completed task using 'approve_task_completion'.
Inputs: requestId (required), taskId (required), completedDetails.
open_task_detailsB
Get details of a specific task by 'taskId'. This is for inspecting task information at any point.
Inputs: taskId (required).
request_planningA
Register a new user request and plan its associated tasks. You must provide 'originalRequest' and 'tasks', and optionally 'splitDetails'.
This tool initiates a new workflow for handling a user's request. The workflow is as follows:
Use 'request_planning' to register a request and its tasks.
After adding tasks, you MUST use 'get_next_task' to retrieve the first task. A progress table will be displayed.
Use 'get_next_task' to retrieve the next uncompleted task.
IMPORTANT: After marking a task as done, the assistant MUST NOT proceed to another task without the user's approval. The user must explicitly approve the completed task using 'approve_task_completion'. A progress table will be displayed before each approval request.
Once a task is approved, you can proceed to 'get_next_task' again to fetch the next pending task.
Repeat this cycle until all tasks are done.
After all tasks are completed (and approved), 'get_next_task' will indicate that all tasks are done and that the request awaits approval for full completion.
The user must then approve the entire request's completion using 'approve_request_completion'. If the user does not approve and wants more tasks, you can again use 'request_planning' to add new tasks and continue the cycle.
The critical point is to always wait for user approval after completing each task and after all tasks are done, wait for request completion approval. Do not proceed automatically.
Inputs: originalRequest (required), tasks (required), splitDetails.
update_taskB
Update an existing task's title and/or description. Only uncompleted tasks can be updated.
A progress table will be displayed showing the updated task information.
Inputs: requestId (required), taskId (required), description, title.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 10 tools
Each tool has a distinct purpose with clear boundaries. For example, 'approve_task_completion' is for approving individual tasks, while 'approve_request_completion' is for finalizing the entire request, and 'mark_task_done' is for marking tasks as done, distinct from approval. There is no overlap that would cause confusion.
All tool names follow a consistent verb_noun pattern with underscores, such as 'add_tasks_to_request', 'approve_request_completion', and 'get_next_task'. The naming is uniform and predictable throughout the set, making it easy to understand each tool's function.
With 10 tools, the server is well-scoped for task management. It covers the full lifecycle of requests and tasks, from creation to completion, without being overly complex or too sparse. Each tool serves a specific role in the workflow, justifying its inclusion.
The tool set provides complete coverage for task management workflows. It includes planning ('request_planning'), task operations (add, update, delete, mark as done, get details, get next), approvals (task and request levels), and listing requests. There are no obvious gaps; agents can handle the entire process from start to finish.
Maintenance
Related MCP Connectors
Headless task management for bots, apps and AI agents. Remote MCP server, OAuth or access token.
Nifty's MCP server — exposes tasks, projects, messages, and files as tools for AI agents.
Project management MCP for AI agents with safe task reads and writes.
Task management your AI agents can actually run. One line becomes a context-ready task over MCP.
Related MCP Servers
- AlicenseAqualityDmaintenanceMCP server for "taming the Claude" with structured task queues.14134 npm71MIT
- AlicenseNot gradedqualityCmaintenanceA Model Context Protocol server that enables advanced task and project management in Todoist via Claude Desktop and other MCP-compatible clients.99 npm2MIT
- FlicenseNot gradedqualityDmaintenanceA task management MCP server that provides tools to create, list, complete, and delete tasks using pluggable storage backends. It enables users to interact with their task lists through natural language using MCP-compatible clients like Claude Desktop.-
- FlicenseBqualityDmaintenanceA simple task management server that allows users to add, list, complete, and delete tasks directly within Claude Desktop. It provides a suite of tools for efficient personal task organization using the Model Context Protocol.4-