MCP Reminder

You can run and manage Xiaozhi MCP alarms and todos through a dedicated MCP server. This service stores your data, parses natural language times, and supports proactive notifications via an MCP endpoint. You can interact with the server using MCP clients to create, query, complete, and monitor alarms and tasks.

How to use

To use this MCP server, connect with an MCP client using either the remote MCP URL or the local runtime configuration. Start the server, then issue commands to add alarms, create todos, check for due items, and receive reminders through your MCP client. The server provides functions to add alarms, list pending items, complete items, and perform comprehensive reminder checks.

How to install

Prerequisites: you need a Python-enabled environment and the UV runtime installed. Ensure you have access to the MCP endpoint you will connect to during runtime.

1. Install UV if you have not already. 2) Clone or download the project files. 3) Prepare your environment by setting the MCP endpoint or using a local runtime configuration.

Additional information

Configuration and running details are provided to help you set up the MCP server for alarms and todos. You will specify the MCP endpoint, use a local command to start the service, and optionally test locally without connecting to the Xiaozhi MCP endpoint.

Key usage patterns include creating alarms and todos via voice or text, querying pending items, and running periodic checks to fetch all due reminders. If you prefer, you can run locally in stdio mode to test the MCP functions without connecting to the remote Xiaozhi endpoint.

Security and identifiers

Protect your MCP endpoint token and any credentials. Do not expose your endpoint URL publicly. Use a secure environment to store tokens and consider rotating credentials as needed.

Troubleshooting and notes

If you encounter connectivity issues, verify that the MCP endpoint is reachable and that your environment allows WebSocket connections. For local testing, confirm that the stdio runtime command is executed with the correct arguments. If reminders do not trigger, ensure the periodic check is configured to run at a suitable interval.

Appendix: Tools and endpoints

The server exposes functions to manage alarms and todos, including adding alarms, querying pending alarms, dismissing alarms, adding todos, listing todos, completing todos, and performing comprehensive reminder checks.

Available tools

add_alarm

Add a new alarm with a time and optional description. Supports natural language time expressions.

get_pending_alarms

Retrieve all alarms that have reached their scheduled time and are not yet dismissed.

dismiss_alarm

Dismiss a specific alarm by its ID.

add_todo

Create a new to-do item with a title, optional remind time, and optional description.

get_pending_todos

Retrieve all todo items that are due or overdue and not yet completed.

complete_todo

Mark a todo as completed, with support for fuzzy title matching.

list_todos

List all todos with optional status filtering (pending, completed, all).

check_all_reminders

Check all alarms and todos for due items and return reminders to be spoken by the AI.