Goal: When I say “Hey Google, talk to my AI assistant” followed by a question, I want Google Home to send that question to my self-hosted n8n server. Then n8n should call the OpenAI API, get the response, and send it back to Google Home so it can speak the answer out loud.
Current status: Nothing is working. Google Home isn’t connecting to my n8n setup at all.
What I’ve completed:
- Created n8n workflow
- Got OpenAI API key
- Set up Google Cloud console account
- Installed SSL certificate using certbot
- Opened firewall ports 80 and 443
- Made n8n accessible from internet
- Partially completed Google’s OAuth2 application form (left many fields empty but it was accepted)
- Generated OAuth2 credentials in n8n
- Stuck on Google asking for “scope” when entering OAuth2 details
Questions:
- Which OAuth2 scope should I specify?
- What exact steps make n8n respond to Google Home voice commands?
Main concern: Is this even possible anymore? I found old tutorials mentioning Dialogflow, but that service ended in June 2023. Can’t locate any current working guides for this setup. Maybe I’m using wrong search terms?
Any help or links to updated tutorials would be appreciated!
That scope issue? Just the beginning unfortunately. Google’s Assistant SDK needs “https://www.googleapis.com/auth/assistant-sdk-prototype” for prototype access, but even with the right scopes, you’ll hit their verification wall that blocks individual developers. I wasted three weeks debugging this exact setup last spring before realizing Google now requires business verification and privacy policy reviews for any Assistant integration handling voice data. Your OAuth app will sit in review limbo forever without a registered business entity. Worse yet, Google killed the Actions SDK for their new Assistant API, but only enterprise partners get access. Your n8n workflow’s probably fine - Google’s the problem. Skip Google entirely. I built a simple Python script with speech recognition that waits for a wake word, then sends audio to my existing n8n webhook. Way better than fighting Google’s red tape and you control everything.
Got this working 6 months ago with a workaround since Google killed direct integrations. Everyone misses this - you need Google Cloud Functions as a bridge between Google Assistant and n8n. Don’t try making Google Home talk directly to n8n. Deploy a simple Cloud Function that grabs voice input from Google Assistant, forwards it to your n8n webhook, waits for OpenAI’s response, then sends it back to Google Home. Skips the OAuth nightmare completely. Flow: Voice → Google Assistant → Cloud Function → n8n → OpenAI → back through the chain. You’ll need Actions API enabled in Google Cloud Console and a basic fulfillment function, but it’s way more reliable than fighting deprecated APIs. Mine’s been stable for months. Latency’s fine - 3-4 seconds total. I can share the basic Cloud Function code for webhook forwarding if you want.
You’re hitting a dead end with Google Home. Dialogflow ES got shut down, then Google replaced it with Actions on Google, which they also killed off for third-party developers in June 2023. That’s why you can’t find current tutorials - Google basically blocked custom voice actions for regular developers.
The OAuth scope “https://www.googleapis.com/auth/actions.fulfillment.conversation” exists but won’t help since the service is deprecated.
Here’s what works: ditch Google Home voice commands entirely. Set up your n8n workflow to respond to webhooks, then use Google Assistant’s “Ask my test app” during development, or pivot to something else.
Try Home Assistant with a voice satellite setup, or just create a simple web interface you can bookmark on your phone. I wasted weeks trying to get Google Home integration working before realizing Google blocked this path for hobbyists.
Your n8n and OpenAI setup sounds solid - just needs a different trigger than Google Home voice commands.
That Google Cloud Functions workaround is clever but way too complex. You’re building a relay system when modern automation platforms already handle this stuff natively.
I hit this exact problem last year when Google killed their third-party voice actions. Wasted weeks trying to patch things together with dead APIs and middleware functions.
What fixed it? Switching to Latenode. They’ve got direct Google Assistant connectors that skip all those OAuth scope headaches you’re dealing with. No Cloud Functions middleman, no fighting n8n’s auth bugs.
Your workflow gets super simple:
Google Assistant → Latenode → OpenAI → response back
Authentication’s automatic. No SSL certs to babysit, no firewall tweaks, no guessing which scopes still work with deprecated services.
Latenode’s OpenAI integration beats anything you’d build in n8n too. When the API hits rate limits or throws errors, it handles retries and fallbacks automatically.
I moved three voice assistant projects from various franken-solutions to Latenode. Everything just works now without constant maintenance.
Check their Google Assistant templates - they’ll show you how to set this up in minutes instead of weeks: https://latenode.com
grab a smart speaker that supports custom integrations. Alexa skills are dead simple compared to Google’s nightmare setup. I switched after days of frustration with this same issue. built my OpenAI integration in 2 hours using Alexa’s developer console - no OAuth pain or broken apis.
Yeah, the Google Home to n8n route is way too complicated for this. Google’s authentication is a nightmare now - they want full app verification for anything production-level.
I’ve built a few AI assistants and there’s a much easier way. Skip Google’s OAuth mess entirely and use something that already handles the API connections.
Latenode cuts through all this BS. It’s got built-in Google Assistant integration without the OAuth headache you’re dealing with. You can wire Google Home straight to OpenAI through their drag-and-drop interface.
Here’s the whole flow:
Google Home → Latenode webhook → OpenAI → back to Google Home
No SSL certs, no firewall tweaks, no guessing which scopes you need.
I ditched n8n for exactly this reason - authentication was always breaking. Latenode just works with Google services out of the box.
If you really want to stick with your current setup, you’d need “https://www.googleapis.com/auth/assistant-sdk-prototype” scope. But honestly? Don’t torture yourself.
Check their Google Assistant templates - it’ll show you how simple this should be: https://latenode.com
Had this exact setup until Google changed their policies this year. Your n8n config isn’t the problem - Google Home Smart Home actions now need Matter compatibility certification for new integrations. Even if you fix the OAuth scope issue, Google will reject your action during review unless you’re a certified Matter device manufacturer. Skip fighting Google’s restrictions and use a Raspberry Pi with a USB mic running Mozilla DeepSpeech instead. Connect it to your n8n workflow via webhooks. The Pi does speech-to-text locally, sends text to n8n, gets your OpenAI response back, then uses espeak or festival for text-to-speech through speakers. This keeps your n8n and OpenAI setup working while ditching Google’s closed ecosystem. Better privacy too since voice processing stays local. Hardware’s cheap and you won’t deal with constant API changes breaking Google integrations.