OpenClaw for Home Assistant: Chat card + Assist agent + automation services (HACS)

[techartdev]

Hi everyone

I’m sharing a new custom integration: OpenClaw for Home Assistant.

GitHub - techartdev/OpenClawHomeAssistantIntegration

This integration connects Home Assistant to OpenClaw and adds:

• Native conversation agent support for Assist
• Lovelace chat card with message history
• Service calls for automations/scripts
• Event hooks for response-driven automations
• Optional voice mode in the chat card - via Browser or HA STT
• Assist Expose-aware context support
• Entities exposed to OpenClaw by the Assist pipeline

If you want to host OpenClaw on Home Assistant device, take a look at the Add-On/App project: OpenClaw/ClawdBot on Home Assistant

This integration uses the OpenClaw OpenAI compatible chat endpoint. In theory it can work with any other OpenAI style API, including Ollama and LM Studio self hosted models.

What it does

• Adds an OpenClaw chat card you can place on any dashboard
• Lets you send messages via service calls from automations
• Fires an event when OpenClaw returns a message
• Supports configurable context behavior (max chars + overflow strategy)
• Supports optional tool-call execution for Home Assistant services
• Supports wake word + always-on voice mode options when using Browser mode (for SST continuous audio is not supported)
• 32 languages supported in Browser voice API

Current highlights

• Improved frontend auto-registration and resource handling
• Better response parsing across different payload formats
• Chat history restore after leaving and returning to dashboard
• Better handling of speech recognition edge cases
• Voice language selection now prioritizes Assist pipeline languages, then falls back on HA configured language

Installation

• HACS installation (recommended), by manually adding the repository
• Restart Home Assistant
• Add OpenClaw from Settings → Devices & Services
• Add the custom card to a dashboard

Notes

• Voice input depends on browser speech APIs; behavior can differ by browser. For Assistant pipelines SST there are limitations.
• If you see browser console errors from other custom cards/extensions, they may be unrelated to OpenClaw
• In Brave Browser, Web Speech is missing or it can be less reliable in future, until properly implemented; there is now an integration option to allow experimental use but currently API is missing and still not working.

Feedback requested

I’d love feedback on:

• Stability across HA versions
• Voice behavior by browser/device
• Context/tool-call defaults
• UX improvements for the chat card

If you try it, please share your HA version, browser, and any logs/screenshots, especially if you experience some issues.
Thanks!

[stephan182730 (Stephan)]

Thx for this plugin. I am not able to get this work with a remote instance of OC. I always get.

The gateway returned an unexpected response — the OpenAI-compatible API is likely disabled. In the OpenClaw addon settings enable ‘enable_openai_api’, restart the addon, and try again.

I already configured HAS so that it can access the remote instance via localhost, but that did not help. I also do not see a way to ‘enable_openai_api’ since this option is not in setup window and and I cant get past the window with the error message.

Stephan

[techartdev]

Hi @stephan182730 , thanks!

Well you should check the docs: OpenClawHomeAssistant/DOCS.md at 6eb2e11862ac995172b206162cf42c5c9d94c843 · techartdev/OpenClawHomeAssistant · GitHub

On your OpenClaw installation, which is remote one (not ran by the OpenClaw HA Addon) you have to enable the API. This can happen by running in terminal: openclaw config set gateway.http.endpoints.chatCompletions.enabled true.
Or by manual edit of the openclaw.json which I don’t advice you to do, as you may accidently break it.

[techartdev]

New updates are released:

Added

• Added HTTPS / SSL support for connecting to OpenClaw gateways running in lan_https mode or behind TLS reverse proxies.
• Auto-discovery now detects access_mode: lan_https and connects to the internal gateway port automatically (no certificate setup needed for local addons).
• Added Verify SSL certificate option in manual config for self-signed certificate environments.
• Added ssl_error config flow error with actionable guidance.
• Added comprehensive remote connection documentation to README with setup table for all access modes.

Fixed

• Fixed “400 Bad Request — plain HTTP request was sent to HTTPS port” when the addon uses lan_https access mode.

Different ways to connect are briefed here: GitHub - techartdev/OpenClawHomeAssistantIntegration

And more detailed info for users who also use the Add-On/App can be found here: OpenClawHomeAssistant/DOCS.md at 6eb2e11862ac995172b206162cf42c5c9d94c843 · techartdev/OpenClawHomeAssistant · GitHub

[techartdev]

New release [0.1.56] is available:

Added

• Event entities (event.openclaw_message_received, event.openclaw_tool_invoked) — native HA EventEntity entities that fire on each assistant reply and tool invocation result. Selectable in the automation UI without YAML.
• Button entities — dashboard-friendly buttons for common actions:
  • Clear History — clears in-memory conversation history
  • Sync History — triggers a backend coordinator refresh
  • Run Diagnostics — fires a connectivity check against the gateway
• Select entity (select.openclaw_active_model) — exposes the list of available models from the gateway’s /v1/models endpoint, allowing model switching from the HA dashboard. Selection is persisted in config entry options.
• Coordinator now caches the full model list (not just the first model) and exposes it via coordinator.available_models.

Now the integration will provide much more user friendly access into automations without the need of raw YAML configure.

If any issues appear, I’m here to help!

[dzmiller]

Connection works great to an external OC. Whats the easiest way to get hass_service_call to work? I would rather not install a third party skill at this point

[techartdev]

Hi @dzmiller !

1. In HA: Settings → Devices & Services → OpenClaw → Configure
2. Turn on Enable tool calls

This will automatically provide OC with tool calls: execute_service / execute_services with a list payload.

OC will then know to use these tools to communicate with HA. OC will send payloads like this:

{
  "list": [
    {
      "domain": "light",
      "service": "turn_on",
      "service_data": { "entity_id": "light.kitchen" }
    }
  ]
}

We currently don’t name the tool as hass_service_call but it behaves the same.

So enabling the tool calls in the configuration is the only thing you have to do.

[dzmiller]

I’m amused by Gerri (open claw) trying to get Jarvis (Assist) to turn on some lights. But I’m still unclear if I need to install one of the HA openclaw plugins if I want to do tool calling from openclaw. I haven’t decided yet how much access I want to give OC to HA. But I do want HA to keep OC up to date on some status info.

I do have tool calling enabled in the integration

[avd706]

I gave my assistant a long lived token for my Homeassistant and it can control everything.

[techartdev]

Hi @dzmiller !

You don’t need the addon if you already have OpenClaw somewhere. You need to enable the tool calls in the settings of the integration so it enables the service calls ability of the integration.

[robertw]

Question where did you store the token and the IP address of the home assistant installation every time I check off and set things up it comes back and says please provide token and the IP address any help would be greatly appreciated thank you

[avd706]

I just told it here is my HA token, it stores it where it stores all the keys.

I just told it to back itself up to GitHub, so every time it starts up I tell it here is a key for GitHub.

[robertw]

thank you but i dont see any sensor data it sees the server just fine

the config is a remote openclaw server api data is fine at the endpoint
http://192.168.1.xx:18789/v1/chat/completions

tested it and see data coming from the gateway…
thankz