YouTip

## Introduction to Oh My OpenAgent (OmO) **Oh My OpenAgent** (commonly abbreviated as **OmO**), formerly known as `oh-my-opencode`, is an open-source AI programming assistant enhancement plugin. OmO runs on top of the **OpenCode** AI development tool, transforming it into a collaborative development team powered by **11 specialized AI Agents**. The core philosophy of OmO is simple: > **You submit a single requirement → OmO automatically orchestrates and dispatches multiple AI agents to work in parallel → Tasks are completed several times faster than with a single AI.** With over **54k+ Stars** on GitHub, OmO is currently the most popular plugin for the OpenCode ecosystem. * **GitHub Repository:** [https://github.com/code-yeongyu/oh-my-openagent](https://github.com/code-yeongyu/oh-my-openagent) * **Official Website:** [https://ohmyopenagent.com/en](https://ohmyopenagent.com/en) --- ## Installation and Setup To install and configure OmO, copy and paste the following prompt directly into your AI coding tool (such as OpenCode, Claude Code, AmpCode, or Cursor): ```text Install and configure oh-my-opencode according to the instructions in this link: https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/refs/heads/dev/docs/guide/installation.md ``` Once executed, the AI will automatically handle the installation process. During setup, you may be prompted to configure your API keys. This step is optional and can be skipped or configured later. --- ## Meet the 11 AI Agents OmO features 11 built-in, specialized agents that work together to manage, plan, write, and review your code. ### Core Agents | Agent | Role | Default Model | When to Use | | :--- | :--- | :--- | :--- | | **Sisyphus** | Main Coordinator & Full-Stack Developer | Claude Opus | The entry point for all tasks; automatically dispatches and coordinates other agents. | | **Prometheus** | Strategic Planner | Claude Opus / GPT-5.4 | Used for planning and structuring complex tasks. | | **Atlas** | To-Do & Task Manager | Kimi K2.5 | Tracks progress across multi-step tasks. | ### Specialist Agents | Agent | Role | Default Model | Key Strengths | | :--- | :--- | :--- | :--- | | **Hephaestus** | Deep Autonomous Worker | GPT-5.3-codex | Long-running, high-intensity independent coding tasks. | | **Oracle** | Architecture Consultant & Debugging Expert | GPT-5.4 | System design, architectural decisions, and troubleshooting complex bugs. | | **Momus** | High-Precision Reviewer | GPT-5.4 | Code reviews and output validation. | | **Metis** | Plan Reviewer | Claude Opus | Audits Prometheus's plans to ensure no edge cases are missed. | ### Utility Agents (Fast & Cost-Effective) | Agent | Role | Default Model | Why Use Lightweight Models? | | :--- | :--- | :--- | :--- | | **Explore** | Codebase Searcher | MiniMax / Grok Code Fast | Code search does not require deep reasoning; speed is the priority. | | **Librarian** | Doc & Code Retriever | MiniMax (Free) | Document retrieval is straightforward and does not require expensive models. | | **Multimodal Looker** | Screenshot & Visual Analyzer | Kimi K2.5 | Kimi excels at multimodal understanding and visual processing. | > 💡 **Pro Tip for Beginners**: You do not need to select these agents manually. **Sisyphus** will automatically analyze your task and dispatch the optimal combination of agents. --- ## Three Work Modes OmO supports three distinct interaction modes depending on the complexity of your task. ### Mode 1: ultrawork (Fully Automated) Add the keyword `ultrawork` or `ulw` to your prompt to trigger full automation: ```text ultrawork help me convert this React component to TypeScript ``` * **Best for:** Everyday development tasks where you want the AI to handle everything autonomously. ### Mode 2: Prometheus Planning Mode (High Precision) Press the **Tab key** to enter Prometheus Mode. The AI will ask clarifying questions to understand your requirements, generate a detailed execution plan, and wait for your approval before writing code. **Workflow:** 1. Press `Tab` to switch to Prometheus Mode. 2. Describe your requirements; the AI will ask clarifying questions. 3. The AI generates a step-by-step execution plan for your review. 4. Type `/start-work` to begin execution. * **Best for:** Complex feature development, architectural refactoring, and tasks requiring precise planning. ### Mode 3: Standard Chat (Flexible) Simply type your question or request without any special keywords: ```text Explain the JWT verification logic in src/auth.ts ``` * **Best for:** General Q&A, code explanations, and minor, single-file modifications. --- ## Configuration File Guide OmO uses two configuration files. If there are conflicting settings, the project-level configuration takes precedence: * **Project-Level Configuration (Highest Priority):** `.opencode/oh-my-opencode.jsonc` * **User-Level Configuration (Global Default):** `~/.config/opencode/oh-my-opencode.jsonc` The configuration files use the JSONC format (JSON with Comments) and support JSON Schema auto-completion: ```jsonc { "$schema": "https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/dev/assets/oh-my-opencode.schema.json", "agents": { // Main Coordinator Agent (Claude Opus series or Kimi is recommended) "sisyphus": { "model": "anthropic/claude-opus-4-6" }, // Architecture Consultant (GPT series is recommended) "oracle": { "model": "openai/gpt-5.4", "variant": "high" } } } ``` > 💡 **Recommendation**: Keep the default settings when you first install OmO. The default configuration is already optimized for most development workflows. --- ## Command & Shortcut Reference ### CLI Commands | Command | Description | | :--- | :--- | | `opencode` | Launch OmO. | | `bunx oh-my-opencode install` | Install or reconfigure the plugin. | | `bunx oh-my-opencode doctor` | Diagnose configuration and environment issues. | | `opencode auth login` | Log in to your AI provider account. | | `opencode models` | List all available models. | | `bunx oh-my-opencode refresh-model-capabilities` | Refresh the cached model capabilities. | ### In-Chat Shortcuts & Keywords | Shortcut / Keyword | Description | | :--- | :--- | | `Tab` | Switch to Prometheus Planning Mode. | | `/start-work` | Begin executing the approved plan in Prometheus Mode. | | `ultrawork` / `ulw` | Include this in your prompt to activate full-scale parallel agent execution. | --- ## Troubleshooting & FAQ ### Q: I ran `opencode` after installation, but I don't see any OmO features. **Solution:** Verify that the plugin is registered correctly in your global configuration: ```bash cat ~/.config/opencode/opencode.json ``` Ensure that the `plugin` array contains `"oh-my-opencode"`. If it is missing, re-run the installer: ```bash bunx oh-my-opencode install ``` ### Q: I see a "Using legacy package name" warning during startup. **Solution:** Update the plugin name from `"oh-my-opencode"` to `"oh-my-openagent"` in your `opencode.json` file: ```json { "plugin": } ``` ### Q: Sisyphus performs poorly and feels like standard Claude. **Cause:** Sisyphus might not be using the Claude Opus model, or you did not trigger the agent workflow. **Solution:** 1. Ensure you are logged into an active Claude Pro/Max account. 2. Explicitly include the `ultrawork` keyword in your prompt. 3. Run `bunx oh-my-opencode doctor` to verify your model routing configuration. ### Q: The task stopped halfway through execution. **Solution:** Simply prompt the AI with: `ultrawork continue the previous task`. Sisyphus will read the state from the Atlas To-Do system and resume from where it left off. ### Q: How do I disable anonymous telemetry? OmO collects anonymous telemetry via PostHog (with hashed data) to improve the tool. If you wish to opt-out, set the following environment variables in your shell profile: ```bash export OMO_SEND_ANONYMOUS_TELEMETRY=0 # or export OMO_DISABLE_POSTHOG=1 ```