OpenClaw + Home Assistant: The Ultimate Smart Home AI Assistant (2026 Guide)

What You'll Get

By the end of this guide, you'll have:

Let's get started.

• **OpenClaw running inside Home Assistant** — no separate server needed
• **Natural language control** — say what you want in plain English
• **Persistent context** — your assistant remembers your preferences and habits
• **AI-powered automations** — leverage LLMs for complex, multi-step tasks
• **Voice integration** — talk to your smart home through Assist

Why OpenClaw + Home Assistant?

Home Assistant is the world's largest open-source smart home platform, with support for over 2,000 device integrations. But it has a learning curve—automations require YAML, triggers, and conditions. That's where OpenClaw comes in.

OpenClaw is an AI agent that:

Together? You get a smart home that *talks back*.

• **Understands natural language** — no syntax required
• **Maintains memory** — learns your habits and preferences
• **Executes multi-step tasks** — chains actions together intelligently
• **Runs 24/7** — always available via Telegram, Discord, or WhatsApp

What Others Are Building

*"I integrated OpenClaw with Home Assistant using ha-mcp, and suddenly my smart home has a brain that persists between conversations. It knows I prefer the thermostat at 72°F when I'm working from home. It knows to turn off the outdoor lights at sunrise. It even reminds me when I've left the garage door open."* — Dan Malone, Home Assistant Community

*"OpenClaw gave my Home Assistant an AI agent with opinions. Now when I ask 'Is everything okay at home?' it doesn't just list device states—it gives me a summary with context: 'Everything's good. The living room is at 71°F, dryer finishes in 12 minutes, and I noticed you left the office window open. Want me to close it?'"*

Prerequisites

Before starting, ensure you have:

**Note:** This guide uses the community-developed OpenClaw Home Assistant add-on. It's not officially maintained by the Home Assistant team, but it's the easiest way to run OpenClaw inside HAOS.

• **Home Assistant OS** running on a supported device (Raspberry Pi 4/5, mini PC, or VM)
• **SSH access** to Home Assistant or the Home Assistant Community Store (HACS)
• **An OpenAI-compatible API key** — OpenAI, Anthropic, or Google Gemini (for the LLM)
• **Basic familiarity** with the Home Assistant interface

Step-by-Step Installation

This section walks you through installing OpenClaw as a Home Assistant add-on. The community-maintained add-on packages everything you need into one installation.

Hardware Requirements

Hardware requirements by device: Raspberry Pi 4 minimum 2GB RAM, recommended 4GB+ RAM. Raspberry Pi 5 minimum 4GB RAM, recommended 8GB RAM. x86 mini PC minimum Intel N100, recommended Intel i3+/AMD Ryzen 3+. VM/Server minimum 2 vCPU with 2GB RAM, recommended 4 vCPU with 4GB RAM.

**Important:** OpenClaw needs around 1-2GB RAM for the LLM context alone. If running other add-ons, budget accordingly.

Repository Setup

The add-on lives in a community repository. Here's how to add it:

• **Open Home Assistant** → Navigate to **Settings** → **Add-ons**
• **Access Add-on Store** → Click the **... (three dots)** in the top-right corner
• **Add Repository** → Select **Repositories** from the dropdown
• **Paste the URL**: https://github.com/techartdev/OpenClawHomeAssistant
• **Confirm** → Click **Add**, then **Close**
• **Refresh** → The Add-on Store should now show "OpenClaw Assistant" at the bottom

Installation Process

Once the repository is added:

• **Find OpenClaw Assistant** → Scroll to the bottom of the Add-on Store
• **Click Install** → This downloads and configures the add-on (2-5 minutes)
• **Review Options** → The add-on comes with sensible defaults, but check: **Start on boot** - enabled by default, **Watchdog** - enabled (auto-restarts if it crashes), **Protection mode** - may need disabled for full functionality

Configuration Deep Dive

The configuration file controls how OpenClaw connects to LLMs and Home Assistant. Here's a complete example:

Configuration options explained: "provider" is the LLM provider (openai, anthropic, google), example: openai. "model" is the model to use, example: gpt-4o or claude-3-5-sonnet. "temperature" controls creativity vs precision (0-1), example: 0.7. "max_tokens" is the maximum response length, example: 2048. "url" is the Home Assistant API endpoint, example: http://supervisor/core. "token" is a long-lived access token.

{
  "llm": {
    "provider": "openai",
    "model": "gpt-4o",
    "api_key": "sk-...",
    "temperature": 0.7,
    "max_tokens": 2048
  },
  "home_assistant": {
    "url": "http://supervisor/core",
    "token": "YOUR_LONG_LIVED_TOKEN"
  },
  "channels": {
    "telegram": {
      "enabled": true,
      "bot_token": "BOT_TOKEN"
    }
  },
  "system": {
    "timezone": "America/New_York",
    "log_level": "info"
  }
}

Getting Your Home Assistant Token

The token is critical—OpenClaw uses it to read sensors and call services:

• **Log into Home Assistant** with your admin account
• **Click your profile** (bottom left, your user icon)
• **Scroll to Long-Lived Access Tokens** (near the bottom)
• **Click Create Token**
• **Name it** something descriptive: "OpenClaw Integration"
• **Copy immediately** — this is the only time you'll see it!
• **Paste into your OpenClaw config**

Starting and Verifying

After configuration:

You should get a list of your Home Assistant entities. If not, check the logs.

• **Click Start** in the OpenClaw add-on card
• **Wait 30-60 seconds** — it needs to initialize the LLM connection
• **Check Logs** — look for "OpenClaw gateway started" or errors
• **Open Web Terminal** — click the "Web UI" link in the add-on Info tab
• **Test with:** What devices do you see in Home Assistant?

Connecting Messaging Channels

OpenClaw inside Home Assistant works great—but adding a messaging channel makes it *sing*. Here are your options:

Option A: Telegram (Recommended)

Create a bot via @BotFather on Telegram, get your **bot token**, and add to OpenClaw config:

{
  "channels": {
    "telegram": {
      "enabled": true,
      "bot_token": "your-telegram-bot-token"
    }
  }
}

Option B: Discord

Create a Discord bot at discord.com/developers, get your **bot token** and **channel ID**, and add to config:

{
  "channels": {
    "discord": {
      "enabled": true,
      "bot_token": "your-discord-bot-token",
      "channel_ids": ["your-channel-id"]
    }
  }
}

Option C: WhatsApp

WhatsApp requires more setup (QR code pairing). See the official OpenClaw WhatsApp guide for detailed instructions.

Practical Use Cases

Now the fun part—what can you actually *do* with this? Here are 10 practical examples, from simple queries to complex automations.

1. Natural Language Device Control

**Instead of:** "Hey Google, turn on the living room lights to 50% and set the thermostat to 72"

**Just say:** "I'm settling in for a movie night"

OpenClaw understands context and executes multiple actions:

# OpenClaw interprets "movie night" as:
- Dim living room lights to 20%
- Close blinds
- Set thermostat to 68°F
- Turn on TV backlight (if connected)
- Disable motion-based automation in living room

2. Smart Context Awareness

**Query:** "Is everything okay at home?"

OpenClaw doesn't just list device states—it provides *intelligent context*:

"Everything looks good. Main floor: 71°F, humidity 45%. The dryer finished 20 minutes ago. I noticed the garage door has been open for 45 minutes—should I close it? Also, the backyard camera detected motion 10 minutes ago (probably the neighbor's cat)."

3. Proactive Alerts

Configure OpenClaw to monitor conditions and alert you:

• **Garage door left open** → "Hey, the garage has been open for 30 minutes"
• **Unusual motion detected** → "Movement detected in the backyard—here's a snapshot"
• **Temperature threshold** → "The greenhouse hit 95°F, I'm turning on the exhaust fan"

4. Complex Automation Queries

**Instead of building complex YAML automations:**

**Just ask OpenClaw:** "Set the thermostat to 72°F on weekdays before I wake up, but only if it's colder than 60°F outside"

OpenClaw translates your intent into Home Assistant actions—and remembers your preferences for next time.

automation:
  trigger:
    - platform: state
      entity_id: binary_sensor.workday_sensor
  condition:
    - condition: time
      after: "06:00:00"
      before: "09:00:00"
  action:
    - service: climate.set_temperature
      target:
        entity_id: climate.thermostat
      data:
        temperature: 72

5. Multi-Step Tasks

**Request:** "I'm heading to work—make sure everything's locked and adjust for my absence"

OpenClaw executes:

• Locks all doors
• Arms security system
• Adjusts thermostat to energy-saving mode
• Turns off all lights
• Activates "away" scene

6. Energy Management

**Query:** "How much energy did we use this week?"

OpenClaw pulls from your energy sensors:

"This week's consumption: 142 kWh, about $18.50 at current rates. That's 12% less than last week. Your solar panels generated 98 kWh, covering 69% of usage. Biggest consumers: HVAC (45%), water heater (22%), entertainment (15%)."

7. Routine Management

**Morning routine:** "Good morning!" → Weather report, today's schedule, thermostat adjustment, coffee maker start.

**Leaving home:** "I'm leaving" → Lights off, thermostat eco-mode, doors locked, security armed.

**Bedtime:** "Good night" → All doors locked, security armed, thermostat night mode, lights dimmed.

8. Device Status Queries

**Query:** "What's using the most power right now?"

"Current draw: 2.4 kW. The HVAC is pulling 1.8 kW (75%), water heater is heating (0.4 kW), and everything else combined is 0.2 kW."

9. Automation Troubleshooting

**Query:** "Why didn't the porch light turn on last night?"

OpenClaw checks logs and explains:

"The motion sensor triggered at 11:47 PM, but the automation was disabled because you put the porch on 'quiet hours' mode last Tuesday. Would you like me to re-enable it, or keep quiet hours active?"

10. Scene Creation

**Request:** "Create a 'date night' scene"

OpenClaw asks clarifying questions, then creates the automation:

"Sure! For date night, I'll set: dim dining lights to 30%, turn on accent lighting, thermostat to 70°F, and play your 'Romance' playlist. Should I also disable motion-triggered announcements?"

Voice Control with Assist

Home Assistant's **Assist** pipeline now integrates with OpenClaw. Here's how:

Now you can walk around your house and have full conversations with your smart home.

• **Enable Assist** in Home Assistant Settings → Voice
• Add the OpenClaw conversation agent
• Push to talk (or use "Hey JARVIS" if you've enabled wake words)

conversation:
  agent: openclaw

Troubleshooting Common Issues

Here's a comprehensive guide to fixing the most common problems users encounter when running OpenClaw with Home Assistant.

Issue: "Failed to connect to Home Assistant"

**Symptoms:** OpenClaw responds with connection errors like "Failed to connect to Home Assistant", "Connection refused", or "Invalid authentication".

**Diagnosis:** Verify your Long-Lived Access Token is valid (it may have expired). Check the token has proper permissions (needs homeassistant read/write). Ensure Home Assistant URL is correct: http://supervisor/core (for HAOS).

**Fix:**

# In Home Assistant, create a new token:
# Profile → Long-Lived Access Tokens → Create Token

# Update your config with the new token and restart the add-on

Issue: "LLM API errors" or "Rate limited"

**Symptoms:** OpenClaw stops responding or returns error messages like "API key invalid", "Rate limit exceeded", or "Insufficient credits".

**Fix:** Check your API key is valid and has credits remaining. Consider switching to a more cost-effective model (see our model comparison guide). Add rate limiting to prevent excessive API calls.

{
  "llm": {
    "provider": "google",
    "model": "gemini-2.0-flash-exp",
    "max_requests_per_minute": 10
  }
}

Issue: Device states not updating

**Symptoms:** OpenClaw reports stale sensor data, or shows "unknown" for device states.

**Diagnosis:** Home Assistant updates states periodically—OpenClaw pulls current states when needed. For real-time updates, ensure your devices report to Home Assistant frequently.

**Fix:**

# Restart the OpenClaw add-on after adding new devices
# This refreshes the entity list

# Also verify entities are properly configured in Home Assistant:
# Settings → Devices & Services → Entities

Issue: Slow responses

**Symptoms:** Commands take 10+ seconds to respond.

**Diagnosis:** LLM API latency (most common cause), slow Home Assistant instance, or too many entities to query.

**Fix:** Use a faster LLM model: **Fastest:** Gemini 1.5 Flash, Claude 3 Haiku. **Balanced:** GPT-4o Mini, Claude 3.5 Sonnet. **Best quality:** GPT-4o, Claude 3 Opus. Run Home Assistant on faster hardware (SSD, more RAM). Reduce the number of devices/sensors OpenClaw needs to query.

{
  "home_assistant": {
    "include_entities": ["climate.*", "light.*", "switch.*"],
    "exclude_entities": ["sensor.*_history", "binary_sensor.*_timestamp"]
  }
}

Issue: "Permission denied" for services

**Symptoms:** OpenClaw can't call certain Home Assistant services like lock.unlock, alarm_control_panel.arm_away, or cover.open_cover.

**Fix:** Home Assistant may restrict certain services by default. Add to your configuration.yaml:

homeassistant:
  allowlist_external_dirs:
    - /config

# Or create a limited API token for OpenClaw
# with only the permissions it needs

Issue: OpenClaw add-on won't start

**Symptoms:** Add-on starts then immediately stops, or shows "Failed to start".

**Diagnosis:** Check the **Log** tab for specific error messages. Common causes: invalid JSON in configuration, missing required fields, or port already in use (18789).

**Fix:**

# Validate your JSON config using a JSON validator
# Common error: trailing commas are not allowed

# Check if port 18789 is in use:
# In Home Assistant terminal: netstat -tuln | grep 18789

Issue: Messages not being received/sent

**Symptoms:** Telegram/Discord messages don't reach OpenClaw, or responses don't come back.

**Diagnosis:** Bot tokens may be incorrect, channel webhook URLs may be wrong, or firewall may be blocking.

**Fix:**

{
  "channels": {
    "telegram": {
      "enabled": true,
      "bot_token": "YOUR_BOT_TOKEN",
      "allowed_telegram_users": ["YOUR_USER_ID"]
    }
  }
}

Issue: Memory/context not persisting

**Symptoms:** OpenClaw doesn't remember previous conversations or preferences.

**Diagnosis:** Check if memory is enabled in config, may be running out of storage, or sessions may be getting cleared.

**Fix:**

{
  "memory": {
    "enabled": true,
    "max_history": 100,
    "persist_directory": "/config/openclaw_memory"
  }
}

Issue: Voice commands not working

**Symptoms:** Assist pipeline doesn't respond to voice commands.

**Fix:** Enable Assist in Home Assistant: **Settings → Voice**. Configure the conversation agent. For wake words, you'll need to set up a separate wake word detector (Porcupine, Snowboy, or the built-in one).

conversation:
  agent: integration:openclaw

Advanced Configuration

Once basic setup works, here are powerful options to unlock more functionality.

Custom Conversation Prompts

Fine-tune how OpenClaw behaves in your home:

{
  "prompts": {
    "system": "You are JARVIS, a helpful smart home assistant for a house in New York. Be concise, proactive, and always confirm dangerous actions like unlocking doors.",
    "greeting": "Good morning! Everything looks good. The current temperature is {sensor.temperature}, and you have {calendar.today's_events} events today."
  }
}

Entity Filtering

Control which devices OpenClaw can see and control:

{
  "home_assistant": {
    "include_entities": [
      "climate.*",
      "light.*",
      "switch.*",
      "lock.*",
      "sensor.temperature_*",
      "sensor.humidity_*"
    ],
    "exclude_entities": [
      "sensor.*_history*",
      "automation.*",
      "scene.*"
    ]
  }
}

Safety Confirmations

Require confirmation for sensitive actions:

{
  "safety": {
    "require_confirmation": ["lock.unlock", "alarm.*", "cover.open_garage"],
    "dangerous_actions_blocked": ["script.delete_.*", "automation.turn_off"]
  }
}

Security Considerations

Running an AI agent with access to your smart home has implications.

Security risks and mitigations: **Unauthorized access** — restrict OpenClaw to specific users (Telegram user IDs, Discord guilds). **API key exposure** — use environment variables, never hardcode keys. **Voice recording** — disable voice if concerned about privacy. **Automation accidents** — start with read-only access, add permissions gradually.

**Important:** The OpenClaw Home Assistant add-on is community-developed. Review the SECURITY.md before production use.

Cost Breakdown

Here's what you'll pay to run OpenClaw + Home Assistant. **Home Assistant** is free (open-source). **LLM API** costs $0.50–$30/month depending on usage. **Hardware** costs $0 (existing HA hardware) or $50–$150 for a Raspberry Pi. **Messaging** is free (Telegram/Discord/WhatsApp).

**Pro tip:** Use Gemini 1.5 Flash or Claude 3 Haiku for simple queries—80%+ cheaper than GPT-4.

Comparison: OpenClaw Inside vs. Outside Home Assistant

Comparison of running OpenClaw inside Home Assistant (as an add-on) vs. outside (standalone): **Setup complexity** — inside is easy, outside is medium. **Hardware requirements** — inside uses HA server, outside needs a separate server/VPS. **Home Assistant integration** — inside is native, outside requires MCP/JSON-RPC. **Resource usage** — inside shares HA resources, outside is independent. **Updates** — inside via HA Add-on store, outside via npm. **Flexibility** — inside is limited to HA capabilities, outside has full OpenClaw features.

**Recommendation:** For most users, the Home Assistant add-on is the easiest path. Power users who want more flexibility should consider running OpenClaw on a separate machine with the MCP Home Assistant integration.

Conclusion

You've got OpenClaw + Home Assistant running. Now what? 1. **Add more skills** — Browse ClawHub for community-built automations 2. **Explore voice** — Enable Assist for hands-free control 3. **Optimize costs** — Switch to cheaper models for simple commands 4. **Secure your setup** — Follow our security hardening guide Related Articles: OpenClaw Setup Guide for Beginners, Best AI Model for OpenClaw in 2026, OpenClaw Security Hardening Checklist, How to Connect OpenClaw to Telegram, OpenClaw Memory System Explained. *Ready to transform your smart home? Start with the add-on installation above, and don't forget to grab your Free Trial on DoneClaw if you want managed hosting with zero DevOps.*

Frequently asked questions