DeerFlow Agent Template

Overview

A starter template for building long-horizon AI agents with DeerFlow, the open-source SuperAgent harness for autonomous project execution. This template demonstrates how to set up sandboxed environments, persistent memory, and sub-agent orchestration for complex multi-step tasks.

Features

• Long-Horizon Configuration: Pre-configured for extended task sessions with context persistence
• Sandboxed Execution: Isolated environments for safe code execution
• Sub-Agent Management: Template for spawning and coordinating specialized sub-agents
• Memory Persistence: Built-in memory system that retains context across sessions
• Tool Integration: Pre-configured tool set for research, coding, and content creation

Quick Start

# Clone DeerFlow
git clone https://github.com/bytedance/deer-flow.git
cd deer-flow

# Copy the template
cp -r templates/research-agent my-agent
cd my-agent

# Configure
cp .env.example .env
# Edit .env with API keys

# Run your agent
python agent.py --task "Your task description"

Template Structure

my-agent/
├── agent.py              # Main agent entry point
├── config.yaml           # Agent configuration
├── skills/
│   ├── research.py       # Research skills
│   ├── coding.py         # Coding skills  
│   └── analysis.py       # Analysis skills
├── memory/
│   └── persistence.py    # Memory configuration
└── tools/
    └── custom_tools.py   # Custom tool definitions

Resources

• DeerFlow GitHub
• DeerFlow Documentation