mcp-notify

The mcp-notify MCP server provides centralized message pushing to a variety of chat and automation platforms (WeCom/WeChat Work, DingTalk, Telegram, Bark, Lark, Feishu, and Home Assistant). It exposes a set of helpers to send text, images, cards, and other notification types through the configured channels. By configuring a single mcp-notify instance, you can route events or prompts to multiple destinations with a consistent interface, enabling your workflows to notify teams and devices across your stack. Tools are organized by target service (e.g., 企业微信群机器人, 企业微信应用号, Telegram Bot) and include functions such as sending text, images, news links, and more, as described in the available tools sections of the README. To use it, start the MCP server with the uvx command (as shown in the configuration), then trigger messages via your MCP client prompts or integration logic that calls the corresponding helper functions (e.g., wework_send_text, tg_send_message, bark_send_notify, etc.).

Prerequisites:

• Python 3.x installed on your system (the uvx-based deployment).
• Network access to allow the MCP server to reach its configured targets (WeWork, DingTalk, Telegram, Bark, Lark, Feishu, Home Assistant, etc.).

Option A — Install and run with uvx (recommended):

1. Prepare a configuration file (e.g., mcp-notify-config.json) with the following content:

{ "mcpServers": { "mcp-notify": { "command": "uvx", "args": ["mcp-notify"], "env": { "WEWORK_BOT_KEY": "your-wework-bot-key" } } } }

2. Ensure Python is installed and accessible in your PATH.
3. Install uvx if necessary (depending on your environment). If uvx is installed as a binary, skip this step.
4. Run the MCP server using the configuration. This may vary depending on how you launch MCP servers in your environment; typically, you would start uvx with the provided config, e.g.: uvx mcp-notify-config.json

Option B — Docker (quick-start):

1. Create a directory for the server and pull up the docker-based setup as described in the README (docker-compose.yml).
2. Start the container: docker-compose up -d
3. Use the provided Docker-based URL in your MCP client to connect to the server (e.g., http://0.0.0.0:8809/mcp for streaming HTTP).

Prerequisites for Docker option: Docker and docker-compose installed on your system.

Note: The README also mentions an alternative Smithery-based integration if you prefer a hosted/serverless approach. You can configure a Smithery URL instead of uvx if that suits your deployment model.

Environment variables: The server supports a wide range of environment variables to configure each target service (WEWORK_BOT_KEY, WEWORK_APP_CORPID, DINGTALK_BOT_KEY, FEISHU_BOT_KEY, LARK_BOT_KEY, TELEGRAM_BOT_TOKEN, HASS_BASE_URL, HASS_ACCESS_TOKEN, etc.). Provide the keys as placeholders if you are setting up a template configuration. Common issues: ensure network access to API endpoints (e.g., WeCom, DingTalk, Telegram), verify bot keys and endpoints, and confirm the target services accept connections from your server. If using the Docker option, ensure the port mappings align with your client configuration (the README references a streaming HTTP endpoint at /mcp). When using Smithery, you’ll need appropriate OAuth or Smithery key per the README. If you encounter authentication or routing problems, double-check the env var values and that the correct mcp-server name (mcp-notify) is being addressed by your client. If you plan to run multiple MCPS, ensure unique server names and distinct configuration blocks.