A custom integration sends voice requests and exposed entities to an n8n workflow that can use AI and trigger automations, speaking the reply back.
A custom integration sends voice requests and exposed entities to an n8n workflow that can use AI and trigger automations, speaking the reply back.
You install the Webhook Conversation custom integration through HACS and point it at the production webhook URL of an n8n webhook node, in a workflow that is set active. Home Assistant sends the conversation history, the user's query and the exposed entities with states, areas and aliases. An n8n agent node handles it and returns JSON, and the reply is spoken. For controlling devices, one user added the Home Assistant call service node.
Users got state and attribute queries working, plus device control through n8n's Home Assistant nodes. Continued conversation reportedly triggers when a reply ends in a question mark. The integration doesn't support tool calls directly, and a 403 error appeared when the webhook required authentication.
The integration times out after 30 seconds by default, so slow multi-tool agent runs fail with an intent recognition error unless you raise the timeout.
Set up an n8n workflow that acts as a voice agent for my Home Assistant at [Home Assistant URL]. It should receive the spoken request and list of exposed entities, use an AI model to decide what to do, call the matching Home Assistant services, and return a short reply to be spoken back. Start with only [list of safe entities, e.g. lights, media player] and exclude locks and alarms; you're done when I can say "[example command]" and it works.
Score every demo request against your ICP, update HubSpot and ping the right rep in Slack within a minute of the form submit.
Every new Zendesk ticket gets a category, priority and sentiment tag plus a draft reply in an internal note before an agent opens it.
Each month, find posts whose search clicks dropped, research what changed with Perplexity, and get a Claude-written refresh brief for each one.