Why switch to fancy new AI browsers when you can use an AI Agent to automate your Chrome?
With fast inference provided by Cerebras for Llama and Qwen models, MCP based tool calls to automate the browser take no time!
This app needs to interact with your local Chrome instance via Chrome Devtools Protocol, so the application has to be run locally.
You need to set up the following environment variables before you can use this in your local environment:
OPENROUTER_KEY, AZURE_SPEECH_KEY, AZURE_SPEECH_REGION
If you are a hackathon judge, kindly reach out to me at akshatbatra25@gmail.com and I'll be quick to provide you values for these variables.
1.) Clone this repo and run npm run install:all
2.) Then npm run start:all
3.) Wait until the AI agent initializes and a list of loaded browser automation tools shows up in the shell/CMD.
Run the following command in CMD/Shell making sure Chrome is in the PATH:
chrome --remote-debugging-port=9000 --remote-allow-origins=localhost
Navigate to localhost:5173 in regular Chrome browser to open the SpeakSurf interface, split screen and start navigating the web in voice mode!
- Open localhost:5173 in your browser.
- Click the microphone button to start voice recognition.
- Speak your browsing instructions.
- Click the microphone button again to stop and send to AI.
- AI will take control of the browser and perform actions as per instructions.
- The final AI response will be played as audio with a pulsating wave animation.