things3-mcp
A minimal MCP server for Things 3 on macOS that enables listing projects and tasks, creating tasks, and updating task status or metadata via natural language.
README
Things MCP
A deliberately small MCP server for the parts of Things used in daily work.
This project is based on hald/things-mcp. It was rebuilt around a much smaller tool and code surface rather than preserving the full upstream feature set.
Tools
list_projects
Lists active Things projects.
Parameters: none.
Each result contains:
id: Project ID used by the other tools.title: Project title.
list_tasks
Lists tasks from Today or from one project.
Parameters:
view(required):todayorproject.project_id: Required whenviewisproject; omit it fortoday.
Each result contains:
id: Task ID used byupdate_task.titlenoteswhen: Scheduled date, when present.statusproject_id:nullfor a standalone task.project: Project title, ornullfor a standalone task.
Examples:
list_tasks(view="today")
list_tasks(view="project", project_id="PROJECT_ID")
create_task
Creates a standalone task or a task inside an existing project.
Parameters:
title(required)project_id: Optional ID returned bylist_projects.when: Optional Things schedule such astoday,tomorrow, orYYYY-MM-DD.notes: Optional task notes.
Examples:
create_task(title="Buy milk", when="today")
create_task(
title="Prepare meeting notes",
project_id="PROJECT_ID",
when="tomorrow",
notes="Cover the launch plan"
)
Things does not return the new task ID through this URL operation. Call
list_tasks afterward when the ID is needed.
update_task
Updates an existing task without moving it or changing unrelated Things data.
Parameters:
task_id(required): ID returned bylist_tasks.title: Replacement title.notes: Replacement notes. Pass an empty string to clear them.when: Replacement schedule.status:open,completed, orcanceled.
At least one change is required. Updating a task requires the Things authorization token described below.
Examples:
update_task(task_id="TASK_ID", notes="Updated context", when="tomorrow")
update_task(task_id="TASK_ID", status="completed")
update_task(task_id="TASK_ID", status="open")
Scope
The server does not expose search, deadlines, tags, areas, headings, checklists, project creation, task movement, bulk operations, or deletion.
Requirements
- macOS with Things 3 installed
- Things URLs enabled in Things Settings
- uv
Updating tasks also requires the Things authorization token. Enable it under Things → Settings → General → Enable Things URLs → Manage.
Enable task updates
Things requires a private authorization token before another app can modify an existing task. The token stays in Things and is read locally by this server; do not paste it into the MCP configuration or share it with an assistant.
On your Mac:
- Open Things.
- Go to Things → Settings → General.
- Enable Things URLs.
- Click Manage next to Things URLs.
- Enable or generate the authorization token.
Creating tasks only requires Things URLs. Editing, completing, canceling, and reopening tasks require the authorization token.
Run
Run directly from GitHub:
uvx --from git+https://github.com/snellingio/things3-mcp things-mcp
Or clone the repository for local development:
git clone https://github.com/snellingio/things3-mcp.git
cd things3-mcp
uv sync --extra test
uv run things-mcp
Example MCP configuration using GitHub:
{
"mcpServers": {
"things": {
"command": "uvx",
"args": [
"--from",
"git+https://github.com/snellingio/things3-mcp",
"things-mcp"
]
}
}
}
Set THINGS_MCP_TRANSPORT=http to use HTTP instead of stdio. The optional
THINGS_MCP_HOST and THINGS_MCP_PORT variables default to 127.0.0.1 and
8000.
Test
uv sync --extra test
uv run pytest
uv run ruff check .
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。