Xiaozhi Mcp Openclaw Official

按小智官方 MCP 接入方式，把小智 AI 设备通过 MCP 接到 OpenClaw / OpenAI-compatible 后端。适用于已经有小智 MCP 接入点（wss://api.xiaozhi.me/mcp/?token=...）的场景。提供一个 `openclaw_query(message)` MCP...

This package is a small bridge that will forward MCP messages to a model backend and return replies. Before installing: (1) Understand that you must provide MCP_ENDPOINT and OPENAI_KEY (these are secrets); the registry metadata incorrectly omits them — treat that as a red flag. (2) Change OPENAI_BASE from the code's default (https://openclaw.994938.xyz/v1) to a backend you trust; leaving the default will send your API key and queries to that third-party host. (3) Treat MCP_ENDPOINT carefully — it likely contains a token in the URL; don't share .env files. (4) Review .env.example and the code locally to confirm behavior and remove any default endpoints you don't trust. (5) Consider running the bridge in a restricted environment/network and monitoring outgoing connections. If you cannot verify the maintainer or the default backend domain, avoid installing or use only with dummy credentials and behind network controls.

Type: OpenClaw Skill Name: xiaozhi-mcp-openclaw-official Version: 1.0.2 The bundle implements a bridge (mcp_pipe.py) that connects a remote WebSocket endpoint (MCP_ENDPOINT) to a local MCP server process, piping stdin/stdout/stderr between them. This architecture allows a remote service (XiaoZhi AI) to drive local tool execution, which is functionally similar to a reverse shell and grants the remote endpoint control over local process communication. While this aligns with the stated purpose of hardware integration, the use of a remote-to-local pipe and a hardcoded third-party backend (https://openclaw.994938.xyz/v1 in openclaw_mcp.py) represents a significant security risk.