Big 3 Super Agent

Big Three Realtime Agents
Voice agent (OpenAI Realtime API) that orchestrates coding agents (Claude Code) and browser agents (Gemini Computer Use)

See this codebase in action here

A unified voice-controlled orchestrator that coordinates three types of AI agents:
1. OpenAI Realtime Voice Agent - Natural voice interactions and orchestration
2. Claude Code Agentic Coder - Software development and file operations
3. Gemini Browser Agent - Web automation and validation

Requirements

- Python 3.11+
- Astral uv - Fast Python package installer and runner
- API Keys: OpenAI, Anthropic (Claude), Google (Gemini)
- Playwright: For browser automation (playwright install after setup)

Install uv if you don't have it:

Setup

1. Clone and Navigate

2. Configure Environment
Copy .env.sample to .env and fill in required values:

3. Install Playwright

4. Run

Architecture

Key Directories & Files

Project Structure

Important Files

- big_three_realtime_agents.py: Main orchestrator script (3000+ lines)
  - Line 184-616: GeminiBrowserAgent class
  - Line 617-1540: ClaudeCodeAgenticCoder class
  - Line 1541-2900: OpenAIRealtimeVoiceAgent class

- Working Directory: apps/content-gen/ (configurable via AGENT_WORKING_DIRECTORY)
  - All Claude Code agents operate with this as their cwd
  - Agents create/modify files relative to this directory
  - Registries stored in agents/ subdirectory

How It Works

1. Voice Orchestration
The OpenAI Realtime Voice Agent acts as the main orchestrator:
- Listens to user voice/text input
- Decides which agent type to use
- Dispatches tasks via tool calls
- Manages agent lifecycle (create, resume, list, delete)

2. Agent Working Directory
Agents are pointed to a specific working directory:

- Claude Code agents: Work in this directory with full file…