Multi-Agent Coder

A multi-agent AI coding system built with LangGraph where specialized agents collaborate through structured debate to solve programming tasks. Includes a VS Code extension for real-time monitoring.

Architecture

The system uses a graph-based workflow with five stages:

1. Research — Gathers context, searches documentation, analyzes requirements
2. Planning — Designs solution architecture and implementation strategy
3. Coding — Generates code with tool-calling agents (OpenAI, Anthropic, or HuggingFace models)
4. Review — Debate framework where agents critique and improve the solution
5. Deploy — Finalizes output with testing and validation

Key Components

• Debate Framework (debate_framework.py) — Structured multi-agent debate for code review and decision-making
• Creativity Engine (creativity_engine.py) — Generates novel approaches and alternative solutions
• HuggingFace Router (hf_routing.py) — Dynamic model selection and routing across providers
• Memory System (memory.py) — Project-aware context with graph-based knowledge storage
• Search Service (search_service.py) — Web search integration for research phase
• Monitoring API (main.py) — FastAPI server exposing real-time workflow state
• VS Code Extension (vscode-extension/) — Real-time dashboard with agent status, progress bars, and activity logs

Tech Stack

• Orchestration: LangGraph (StateGraph with checkpointing)
• LLMs: OpenAI, Anthropic, HuggingFace (configurable routing)
• Framework: LangChain (tool-calling agents, prompt templates)
• API: FastAPI + Uvicorn
• Monitoring UI: VS Code extension (TypeScript) + Next.js dashboard
• Testing: pytest

Project Structure

multi-agent-coder/
├── agents.py                # LangGraph agents with tool-calling
├── graph.py                 # StateGraph workflow (research → plan → code → review → deploy)
├── debate_framework.py      # Multi-agent debate for code review
├── creativity_engine.py     # Novel approach generation
├── hf_routing.py            # HuggingFace model router
├── memory.py                # Graph-based project memory
├── search_service.py        # Web search integration
├── main.py                  # FastAPI monitoring server + CLI entry point
├── tests/                   # Test suite
└── vscode-extension/        # VS Code monitoring extension
    ├── src/extension.ts      # Extension entry point
    ├── app/                  # Next.js dashboard UI
    └── components/           # UI components

Quick Start

git clone https://github.com/paxtonedgar/multi-agent-coder.git
cd multi-agent-coder

pip install -r requirements.txt

# Set API keys
export OPENAI_API_KEY=your_key
# or
export ANTHROPIC_API_KEY=your_key

# Run a task
python main.py --task "Build a REST API for user management"

# Or start the monitoring server
python main.py --serve

VS Code Extension

cd vscode-extension
pnpm install
# Press F5 in VS Code to launch extension dev host

Opens a real-time dashboard (Ctrl+Shift+A / Cmd+Shift+A) showing agent status, progress, and activity logs.

Testing

pytest
pytest test_enhanced_search.py -v

License

MIT