Oh My Openagent
omo; the best agent harness - previously oh-my-opencode
Last verified:
What is Oh My Openagent?
Oh My OpenAgent is a multi-model agent orchestration harness for OpenCode that turns a single AI agent into a coordinated development team. It is designed for CLI-based coding workflows and emphasizes planning, delegation, verification, and parallel execution rather than a single model doing everything alone.
Its core workflow centers on specialized agents such as Sisyphus, Prometheus, Atlas, Oracle, Librarian, Explore, Metis, and Momus. The site describes Sisyphus as the main orchestrator, Prometheus as the strategic planner, and Atlas as the executor that distributes work to specialized agents and verifies outcomes independently. It also highlights IntentGate, which classifies what you actually mean before acting.
The product is built for developers who want more autonomous coding help on complex tasks like feature work, refactoring, debugging, codebase exploration, and documentation search. The website positions it as useful for people who want better orchestration across models, cheaper task routing, and less manual prompt wrangling while still keeping control through planning modes and configuration.
A big emphasis is placed on autonomy and continuity. The site says it supports ultrawork mode, background agents, session continuity via boulder.json, and independent verification so interrupted work can resume where it stopped. It also advertises specialized tools like LSP and AST-aware code operations, lifecycle hooks, and team mode for parallel specialist agents.
Oh My Openagent pricing
Pricing model: Freemium
The website presents Oh My OpenAgent as free to install and use, with the install command shown as bunx oh-my-openagent install. It does not show a paid subscription or tiered pricing on the main site or docs page. The site instead emphasizes open-source use, provider-based model access, and configuration flexibility, with costs determined more by the underlying model providers you connect than by the harness itself.
Oh My Openagent pros
- Multi-model orchestration
- Specialized agent roles
- Parallel background agents
- Independent verification
- Session continuity
- Intent-based routing
- Ultrawork automation
- Prometheus planning mode
- Atlas execution mode
- Oracle architecture help
- Librarian docs search
- Explore codebase grep
- Metis gap analysis
- Momus plan review
- Team Mode parallelism
- 54+ lifecycle hooks
- LSP and AST tools
- Custom agent creation
- Model fallback chains
- No vendor lock-in
Oh My Openagent cons
- Complex setup compared with simple chat tools
- Best value depends on OpenCode
- Some advanced features require configuration
- Team Mode is opt-in and not default
- Many features depend on model availability
- Can be overkill for simple tasks
- Multi-agent debugging may be difficult
- Outputs still depend on provider behavior
Frequently asked questions about Oh My Openagent
What is Oh My OpenAgent?
Oh My OpenAgent is a multi-model agent orchestration harness for OpenCode. It takes a single AI agent workflow and turns it into a coordinated system of specialized agents that plan, delegate, execute, and verify coding tasks.
What does Sisyphus do?
Sisyphus is the main orchestrator. It parses intent, assesses the codebase, routes work to specialist agents, and performs independent verification so the system can keep moving even on long or interrupted tasks.
How does ultrawork mode work?
Ultrawork mode is the hands-off command path described on the site. You type ultrawork or ulw, and the system handles planning, exploration, implementation, verification, and self-correction without needing step-by-step manual guidance.
What is Prometheus mode for?
Prometheus is the strategic planning agent. It interviews you, asks clarifying questions, identifies ambiguities, and builds a detailed plan before any code is written, which is useful for larger or more complex work.
What does Atlas do?
Atlas is the executor. It takes verified plans, distributes work to specialized subagents, accumulates learnings across tasks, and independently verifies completion rather than trusting subagent output blindly.
What is Team Mode?
Team Mode enables multiple specialist agents to work in parallel under a lead agent. The site says it is opt-in, uses dedicated team_* tools, supports real-time messaging, and can run up to 8 members simultaneously.
Which agents are built in?
The site lists built-in agents including Sisyphus, Prometheus, Atlas, Oracle, Librarian, Explore, Metis, and Momus. It also describes a dynamic agent system that can assemble custom agents on demand based on task category and skill needs.
Does it support different models?
Yes. The site explicitly emphasizes multi-model orchestration and says it can route work across models such as Claude, GPT, Gemini, Kimi, and others depending on the category and task type.
How does session continuity work?
The site says work is tracked in boulder.json so sessions can resume after interruptions such as crashes or power loss. The idea is that you can pick up exactly where you left off without losing task state.
How do I start using it?
The website shows a simple install command, bunx oh-my-openagent install, and the docs say you can also paste an installation instruction into an LLM agent session. After installation, the site says you can start with ultrawork and let the system figure out the rest.