_{Chrome Workflows Extension v0.5.0}

Local, private AI automation with Chrome's built-in Gemini Nano and n8n workflows.

This project connects n8n (workflow automation) with Chrome's built-in AI (Gemini Nano), allowing you to create AI-powered workflows that run entirely on your device.

A Node.js server with a web interface that bridges n8n and Chrome's AI APIs.

• Runs on localhost:3333
• Must be open in a Chrome tab
• Processes AI requests from n8n
• 100% local, private AI

Custom n8n nodes that expose Chrome's AI capabilities:

• Chrome Prompt AI ✅ - General text generation (works reliably)
• Chrome Writer - Text writing with tone/format
• Chrome Summarizer - Text summarisation
• Chrome Translator - Language translation
• Chrome Rewriter - Text rewriting
• Chrome Proofreader - Grammar checking
• Chrome Language Detector - Language detection

Note: Currently, only Prompt AI works reliably. Other APIs are experimental and require user interaction.

Browser extension to trigger n8n workflows from any webpage via right-click.

• Trigger workflows with context menu
• Pass page data to workflows
• Manage workflows from extension

1. Chrome with AI enabled:

   • Join Chrome AI Early Preview
   • Enable flags: chrome://flags/#prompt-api-for-gemini-nano
   • Download model: chrome://components/ → "Optimization Guide On Device Model"

2. n8n installed:

   npm install -g n8n

1. Install Chrome AI Nodes:

   cd packages/n8n-nodes-chrome-ai
   npm install && npm run build && npm link
   cd ~/.n8n/custom
   npm link n8n-nodes-chrome-ai

2. Start the Bridge Server:

   cd webapp
   npm install
   npm start

   Open http://localhost:3333 in Chrome and keep it open.

3. Start n8n:

   n8n start

   Open http://localhost:5678

4. Configure n8n:

   • Go to Settings → Credentials
   • Add "Chrome AI API" credential
   • Set Bridge URL: http://localhost:3333

5. Create your first workflow:

   • Add "Chrome Prompt AI" node
   • Configure prompt and parameters
   • Run the workflow!

./install.sh

./start-platform.sh

Starts the bridge server and n8n.

./stop-platform.sh

Stops all services.

All what you need to do is to drag and drop or use the import file on the top right menu and select the Json file to a N8n canvas to run them.

These workflows are the same once that been showcased in the demo.

I recommend starting with the Chat with Chrome Prompt AI workflow, because doesn't need any crendentials to setup.

• ✅ 100% Local - All AI processing on your device
• ✅ Private - No data sent to external servers
• ✅ Free - No API costs, uses Chrome's built-in AI
• ✅ Fast - On-device inference, no network latency
• ✅ Integrated - Works seamlessly with n8n workflows
• ✅ Session Management - Continue conversations across workflow runs
• ✅ Fallback Support - Automatically falls back to Prompt AI when specific APIs are unavailable
• ✅ Concurrent Sessions - Multiple AI operations can run simultaneously

• Web app must be open in a Chrome tab
• Chrome AI flags must be enabled
• Gemini Nano model must be downloaded (~1.5GB)
• All Chrome AI nodes now work reliably with automatic fallback to Prompt AI
• Writer API may still require user interaction, but has fallback support

• QUICKSTART.md - Get started in 5 minutes
• docs/ARCHITECTURE.md - System architecture
• docs/NODE-REFERENCE.md - n8n nodes reference
• docs/TROUBLESHOOTING.md - Common issues and fixes
• webapp/README.md - Bridge server details
• chrome-workflows/readme.md - Extension details

• Content Generation: Create articles, emails, social posts
• Data Processing: Summarize documents, extract information
• Translation: Translate text between languages
• Code Assistance: Generate code snippets, explain code
• Research: Analyse text, answer questions
• Automation: Any workflow that needs AI text processing

• Chrome: Version 127+ with AI features
• n8n: Any recent version
• Node.js: Version 16+
• Operating System: macOS, Linux, or Windows

• All AI processing happens locally on your device
• No data is sent to external servers
• Gemini Nano runs on-device
• Communication is only on localhost
• Your data never leaves your machine

1. Model download is large - Gemini Nano is ~1.5GB
2. Chrome flags required - Must enable experimental features
3. Single tab requirement - Web app must stay open
4. Writer API user gestures - May require user interaction but has fallback support

Keep the web app open in Chrome for workflows to work!