This guide provides a rapid getting-started path for oh-my-openagent (formerly oh-my-opencode). It covers installation via the CLI and executing your first ultrawork command to engage the multi-agent orchestration system.
The setup process transitions from package installation to an interactive configuration phase that auto-detects your AI subscriptions (Claude, OpenAI, Gemini, etc.) to map the best "brains" to specific agent roles docs/guide/installation.md3-6 The system uses a multi-model approach where different providers are orchestrated based on their specific reasoning strengths—for example, Claude-family models for orchestration (Sisyphus) and GPT-family models for deep technical execution (Hephaestus) docs/guide/agent-model-matching.md13-37
The following diagram bridges the user's intent to the underlying code entities responsible for initialization and execution.
Sources: src/cli/cli-installer.ts30 src/cli/cli-program.ts17 docs/reference/features.md13 docs/guide/overview.md29-32
The recommended way to get started is using the interactive installer. This handles plugin registration and model mapping based on your available API keys.
Paste this prompt into your existing LLM agent (Claude Code, Cursor, etc.):
If you prefer manual control, run the installer directly. The CLI supports both TUI and non-interactive modes docs/guide/installation.md35-38
What the installer does:
oh-my-openagent.jsonc docs/reference/configuration.md48-50opencode.json configuration docs/guide/installation.md12Sources: docs/guide/installation.md3-14 docs/reference/configuration.md44-59
Launch OpenCode to trigger the plugin initialization sequence. The CLI provides a models command to verify which models are successfully resolved and available for use docs/guide/agent-model-matching.md98-99
The installation and configuration sequence manages the setup of the orchestration environment:
Sources: docs/guide/installation.md12-14 docs/guide/agent-model-matching.md98-99 docs/reference/configuration.md15
ultraworkOnce inside the OpenCode prompt, type the ultrawork (or ulw) command. This is the "autopilot" mode where the system engages all agents to execute until task completion docs/guide/overview.md29-32
Sisyphus orchestrator (optimized for Claude Opus 4.7 or Kimi K2.6) to plan, delegate, and execute complex tasks with aggressive parallel execution docs/reference/features.md13Hephaestus (optimized for GPT-5.5) for deep coding or Oracle for architecture consultation docs/reference/features.md14-15The system behavior is governed by a JSONC configuration file (e.g., oh-my-openagent.jsonc) that defines agents, categories, and provider settings docs/reference/configuration.md75-142
| Feature | Description | Config Section |
|---|---|---|
| Agents | Multi-agent definitions and model overrides | agents docs/reference/configuration.md79-97 |
| Categories | Task-based model routing (e.g., visual-engineering) | categories docs/reference/configuration.md99-124 |
| Concurrency | Limits for providers and specific models | background_task docs/reference/configuration.md127-138 |
| Experimental | Feature flags like task_system or aggressive_truncation | experimental docs/reference/configuration.md140 |
Sources: docs/guide/overview.md29-32 docs/reference/features.md11-32 docs/reference/configuration.md71-143 docs/guide/agent-model-matching.md13-16
| Symptom | Cause | Fix |
|---|---|---|
Command not found: omo | CLI not in PATH or not installed via bunx. | Always invoke via bunx oh-my-opencode or use the alias lazycodex-ai docs/guide/installation.md16 |
Sisyphus is failing | Unsupported model used for orchestrator. | Ensure Sisyphus is pointed at Claude (Opus 4.7), Kimi (K2.6), GLM (5.1), or GPT (5.5) docs/guide/agent-model-matching.md11-18 |
Hephaestus won't activate | Missing GPT-5.5 capable provider. | Hephaestus requires a GPT-capable provider (OpenAI, Copilot, etc.) docs/reference/features.md14 |
Codex install failed | Missing Git Bash on Windows. | Run the installer again; it will attempt a winget install of Git docs/guide/installation.md43-48 |
Sources: docs/guide/installation.md16-60 docs/guide/agent-model-matching.md11-18 docs/reference/features.md14
Refresh this wiki