Browser Agent CLI

AI-powered browser automation with voice commands. Uses Playwright MCP for browser control, Claude for interpreting commands, and ElevenLabs for text-to-speech.

Prerequisites

• Node.js (for running Playwright MCP server)
• Python 3.11+
• ffmpeg (for audio playback in voice mode)

Setup

1. Install dependencies:

uv sync

2. Install Playwright MCP globally:

npm install -g @playwright/mcp

3. Set your API keys:

export ANTHROPIC_API_KEY=your-key-here
export OPENAI_API_KEY=your-key-here      # For Whisper speech-to-text
export ELEVENLABS_API_KEY=your-key-here  # For text-to-speech

Or create a .env file with these values.

Usage

Text Mode (default)

uv run python -m browser_agent https://example.com

Voice Mode

uv run python -m browser_agent --voice https://example.com

In voice mode:

• Press Enter to start recording your voice command
• Press Enter again to stop recording and submit
• The agent will speak what it's doing and its responses

Example Commands

> Click the login button
> Fill in the email field with test@example.com
> Scroll down to the footer
> Take a screenshot
> What text is on this page?

Exiting

Type quit or exit to close the browser and exit.

Features

• Voice input: Speak commands using your microphone (Whisper transcription)
• Voice output: Agent speaks what it's doing (ElevenLabs TTS)
• Natural language commands: Just describe what you want the browser to do
• Full browser control: Click, type, scroll, navigate, screenshot, and more
• Conversation history: Claude remembers context from previous commands