Tutorial

Oh My OpenCode Quick Start & Best Practices

Learn how to quickly get started with Oh My OpenCode, configure multi-agent systems, and follow best practices.

1. What is Oh My OpenCode?

Oh My OpenCode is an open-source agent orchestration framework for OpenCode. Its main features include:
  • Multi-Agent Architecture
    • Sisyphus: the main agent responsible for task planning and execution
    • Librarian: documentation and knowledge retrieval
    • Oracle: reasoning and suggestions
    • Explore: exploratory tasks
    • Other agents: frontend assistance, documentation, multimodal processing
  • Automated Task Execution Agents can independently or collaboratively complete complex tasks, reducing manual intervention.
  • LSP & AST Support Supports Language Server Protocol (LSP) and Abstract Syntax Tree (AST) operations for precise code modifications and refactoring.

2. Quick Installation

Assuming you already have OpenCode installed:
# Clone the Oh My OpenCode repository
git clone https://github.com/code-yeongyu/oh-my-opencode.git
cd oh-my-opencode

# Install dependencies
pip install -r requirements.txt
⚠️ Note: Python >= 3.10 is required. A virtual environment is recommended.

3. MCP & Agent Configuration

  1. Configure MCP Server
Edit config/mcp.yml:
agents:
  - name: Sisyphus
    role: main
    model: gpt-4
  - name: librarian
    role: doc-helper
    model: gpt-3.5
  1. Set Environment Variables
export OPENAI_API_KEY="YOUR_API_KEY"
export OH_MY_OPENCODE_MODE="production"
  1. Enable Claude Compatibility (Optional)
compatibility:
  claude: true

4. Running Your First Agent

python run_agent.py --agent Sisyphus
  • Runs the default example task
  • Use --task "Generate README" to run a custom task
Tip: Use --debug to see each step executed by the agent for easier troubleshooting.

5. Best Practices

  1. Prioritize Task Decomposition
    • Break down complex tasks before assigning them to Sisyphus or other agents.
  2. Configure Multi-Agent Setup Wisely
    • Adjust agent number and roles based on task complexity
    • For small tasks, only Sisyphus and Librarian may suffice
  3. Combine Local & Cloud Models
    • Local Ollama models for private data
    • Cloud models for heavy reasoning tasks
  4. Check Logs Regularly
    • Complex task execution may occasionally need manual intervention
  5. Version Control
    • Use git to manage configs and scripts to ensure reproducibility

6. Further References

  • Multi-Agent Example Tasks: Oh My OpenCode Examples
  • MCP Advanced Configuration Guide: /en/mcp
  • OpenCode CLI Integration: /en/cli-commands

7. Summary

Oh My OpenCode is not just an “AI code generator”; it is a fully orchestrated multi-agent execution platform. By mastering MCP configuration, agent roles, and task decomposition, you can automate complex coding workflows within the OpenCode ecosystem.