AgentForge

AgentForge automatically creates complete, immediately runnable multi-agent research systems based on your research project requirements.

Usage

1. Fill in User Input

Open user_input.md and fill in your research project information as prompted:

• Research topic and core questions
• Data collection requirements
• Analysis methods
• Report format and style

If you're unsure about certain questions, you can fill in "Suggested by AI".

2. Invoke the Generator

Execute in Claude Code:

Please read the information I filled in user_input.md,
then use the generator.md prompt,
to generate a research multi-agent system for me.

3. Obtain the Generated System

The generator will create a complete project for you, including:

{your-project-name}/
├── agents/                  # 4 agent blueprints
├── data/                    # Data directories (3 phases)
├── references/              # Reference materials
├── wip/                     # Work-in-progress notes
└── README.md                # Project documentation

4. Run the System

After generation, execute:

Please use agents/00.orchestrator.md to start executing the research task

The system will automatically complete: data collection → analysis → report generation

Generated System Architecture

Based on declarative multi-agent architecture, using five core patterns:

1. Prompt-Defined Agents - Define agent behavior with natural language
2. Orchestrated Pipeline - Three-phase sequential execution, data refined progressively
3. Filesystem Data Bus - Data passed through files, completely transparent
4. Parallel Instance Execution - Support parallel processing of multiple data sources
5. Reference Data Configuration - Domain knowledge externalized as independent files

For detailed architecture description, see generator.md.

Suitable Scenarios

Particularly suitable for:

• Information-intensive research (extensive data collection and analysis)
• Periodic research (repeatable execution)
• Quality-first research (transparent and traceable)

File Descriptions

• user_input.md - User fills in research project information
• generator.md - Generator prompt (includes architecture principles and generation logic)
• README.md - This file (usage instructions)

---

Quick start: First fill in user_input.md, then invoke the generator!