# Home Assistant MCP — Control Home Assistant from Agents

> Home Assistant MCP exposes smart-home actions as MCP tools so agents can query state and trigger automations via a controlled, auditable interface.

## Install

Merge the JSON below into your `.mcp.json`:

## Quick Use

1. Install / run:
   ```bash
   git clone https://github.com/homeassistant-ai/ha-mcp && cd ha-mcp
   ```
2. Start / smoke test:
   ```bash
   sed -n '1,260p' README.md
   ```
3. Verify:
   - Start the server as documented, then from your MCP client call one read-only tool (e.g., list entities) before you enable any write actions.

## Intro

Home Assistant MCP exposes smart-home actions as MCP tools so agents can query state and trigger automations via a controlled, auditable interface.

- **Best for:** Home Assistant users who want agent-driven status checks and automations without giving the agent full UI access
- **Works with:** An MCP-capable agent client + a Home Assistant instance (credentials per repo docs)
- **Setup time:** 20 minutes

## Practical Notes

- Setup time ~20 minutes (credentials + run server + connect client)
- Safety check: keep write actions disabled until read-only calls are stable
- GitHub stars + forks (verified): see Source & Thanks

Treat smart-home tool calls like production ops. Define ‘safe defaults’ (read-only), then add a small set of allowed writes with guardrails. If your agent can’t explain why it’s turning something on, it shouldn’t be allowed to do it.

### FAQ

**Q: Can I let an agent control everything?**
A: You can, but start with read-only tools and explicitly whitelist safe actions.

**Q: How do I prevent accidental triggers?**
A: Require confirmations for writes, and add rate limits / time windows for automations.

**Q: Is this better than UI automation?**
A: Yes for reliability: tool calls are structured and auditable, unlike fragile UI clicks.

## Source & Thanks

> Source: https://github.com/homeassistant-ai/ha-mcp
> License: MIT
> GitHub stars: 2,807 · forks: 114

---

<!-- ZH -->

## 快速使用

1. 安装 / 运行：
   ```bash
   git clone https://github.com/homeassistant-ai/ha-mcp && cd ha-mcp
   ```
2. 启动 / 冒烟测试：
   ```bash
   sed -n '1,260p' README.md
   ```
3. 验证：
   - 按文档启动 server 后，先从 MCP 客户端调用一个只读工具（如列出实体），确认无误再开启写入类动作。

## 简介

Home Assistant MCP 把智能家居的“查询状态/触发自动化”能力封装成 MCP 工具，让 agent 以最小权限、可审计的方式调用，并把家庭控制从 UI 自动化升级为结构化工具调用。

- **适合谁：** Home Assistant 用户：想让 agent 做状态查询/触发自动化，但不想开放完整 UI 权限
- **可搭配：** 支持 MCP 的客户端 + Home Assistant 实例（凭证配置以仓库文档为准）
- **准备时间：** 20 分钟

## 实战建议

- 上手约 20 分钟（配置凭证 + 启动 server + 连接客户端）
- 安全验收：只读调用稳定前，不要开放写入类动作
- GitHub stars + forks（已核验）：见「来源与感谢」

把智能家居的工具调用当成“生产运维”来做：默认只读，再加少量可写动作并设置护栏。如果 agent 不能清楚解释为何要开灯/开空调，那它就不该被允许这么做。

### FAQ

**能让 agent 控制一切吗？**
A: 技术上可以，但建议先只读，再按白名单逐步开放安全动作。

**如何避免误触发？**
A: 写入类动作要求二次确认，并加上频控与时间窗口限制。

**比 UI 自动化更好吗？**
A: 在可靠性上更好：工具调用结构化、可审计，不像 UI 点击那么脆弱。

## 来源与感谢

> Source: https://github.com/homeassistant-ai/ha-mcp
> License: MIT
> GitHub stars: 2,807 · forks: 114


---
Source: https://tokrepo.com/en/workflows/home-assistant-mcp-control-home-assistant-from-agents
Author: MCP Hub