Agent Network Expander Template

A tool for analyzing and expanding multi-agent AI systems by identifying gaps and suggesting complementary orchestration and action agents through an iterative, interactive approval workflow.

Overview

This repository provides tooling to review existing multi-agent network configurations and intelligently suggest additional agents to enhance coordination, coverage, and capability.

Two Operating Modes

1. Iterative Mode (Recommended): Suggests agents one at a time, allowing you to approve/reject each suggestion individually
2. Batch Mode: Analyzes the entire network and generates all suggestions at once

📊 See MODE_COMPARISON.md for detailed comparison and recommendations.

Two Agent Layers

1. Orchestration & Routing Layer: Agents that coordinate workflows, manage priorities, route tasks, and handle system-level concerns
2. Action Layer: Specialized agents that perform specific tasks and operations

What It Does

The Agent Network Expander analyzes your multi-agent system and:

• Maps existing agents and their relationships
• Identifies functional gaps and coordination needs
• Suggests complementary orchestration agents for better workflow management
• Recommends specialized action agents to fill capability gaps
• Generates detailed analysis documents with implementation priorities
• Optionally creates system prompts for recommended agents

Usage

Iterative Mode (Recommended)

Claude Code (Cloud)

# In your agent network directory
/review-agent-network-iterative

The command will:

1. Analyze your agent configuration
2. Present agent suggestions one at a time
3. Ask for your approval on each suggestion
4. Generate system prompts for accepted agents
5. Create a summary of all decisions

Local (Ollama)

# Create and activate virtual environment
python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Run iterative analysis
python scripts/analyze_network_iterative.py /path/to/agent/config

# Or analyze current directory
python scripts/analyze_network_iterative.py .

Iterative workflow:

• Review each suggestion individually
• Choose: Yes (add), No (reject), Skip (maybe later), or Quit (stop reviewing)
• System prompts generated immediately for accepted agents
• Summary saved with all decisions

Batch Mode

Claude Code (Cloud)

/review-agent-network

Local (Ollama)

python scripts/analyze_network_local.py /path/to/agent/config

Batch workflow:

• Generates complete analysis document
• Lists all suggestions at once
• Optionally generates all system prompts in bulk
• Best for comprehensive planning sessions

Deployment Options

1. As a Slash Command in Claude Code

• Iterative: Copy slash-commands/review-agent-network-iterative.md to .claude/commands/
• Batch: Copy slash-commands/review-agent-network.md to .claude/commands/
• Restart or reload Claude Code
• Access via /review-agent-network-iterative or /review-agent-network

2. As a Standalone CLI Tool

• Install the script globally or add to your PATH
• Use as a command-line utility for batch processing multiple agent networks
• Integrate into CI/CD pipelines for continuous agent architecture review

3. As a GitHub Action

• Schedule periodic reviews of agent network configurations
• Automatically generate pull requests with suggestions
• Track agent network evolution over time

4. As a Local Development Tool

• Keep in your tools directory (~/programs/agent-network-expander/)
• Use during agent system design and development
• Run analysis before major architecture changes

5. As Part of Agent Development Workflow

• Integrate into agent creation templates
• Run during agent network initialization
• Use for periodic architecture health checks

Repository Structure

.
├── slash-commands/
│   ├── review-agent-network-iterative.md  # Interactive review mode
│   └── review-agent-network.md            # Batch analysis mode
├── scripts/
│   ├── analyze_network_iterative.py       # Local iterative analyzer
│   └── analyze_network_local.py           # Local batch analyzer
├── findings/                              # Generated analysis outputs
│   ├── analysis-summary.md                # Iterative mode summary
│   ├── agent-network-analysis.md          # Batch mode analysis
│   └── system-prompts/
│       ├── orchestration/
│       └── action/
└── requirements.txt                       # Python dependencies

Requirements

For Claude Code Version

• Claude Code CLI
• Active Anthropic API access

For Local Version

• Python 3.8+
• Ollama installed and running
• Recommended model: qwen2.5:14b-instruct-q5_K_M (or similar 10GB+ model)

Use Cases

• Initial Architecture Design: Identify what agents you need before building
• System Expansion: Find gaps in existing agent networks
• Optimization: Discover redundancies and missing coordination
• Documentation: Generate comprehensive agent network maps
• Team Onboarding: Help new developers understand agent architecture

Contributing

Contributions welcome! This template is designed to be extended with:

• Additional analysis heuristics
• Support for different agent frameworks
• Alternative LLM backends
• Visualization tools
• Integration with agent orchestration platforms

License

MIT License - See LICENSE file for details.

Author

Created by Daniel Rosehill

• Website: danielrosehill.com
• Email: [email protected]