Oh My OpenAgent (OmO) / oh-my-opencode¶

What it is¶

Oh My OpenAgent (previously oh-my-opencode) is an open-source agent harness designed to provide a "world-class" development experience across multiple LLM providers. It acts as an orchestration layer on top of OpenCode, offering a full AI developer team in your terminal.

What problem it solves¶

It solves the "harness problem" where models fail not because of intelligence, but because of poor edit tools and narrow context. It provides reliable multi-model orchestration, surgical editing via content hashes, and a "Discipline Agent" system that ensures tasks are driven to 100% completion.

Where it fits in the stack¶

Development & Ops / Agent Harness. It is the open-source alternative to proprietary "walled garden" agents like Claude Code or AmpCode.

Typical use cases¶

Complex Feature Building: Using the ultrawork command to trigger a multi-agent plan-and-execute loop.
Large-scale Refactoring: Leveraging LSP and AST-Grep integration for deterministic code changes.
Autonomous Debugging: Deploying specialized agents (Oracle, Librarian) to root cause and fix elusive bugs.
Browser Automation: Using the built-in Playwright skill for UI testing or data scraping.

Key Agents¶

OmO uses a "Discipline Agent" system where specialized agents collaborate: - Sisyphus: The main orchestrator. Plans, delegates, and ensures tasks never stop halfway. - Hephaestus: The "Deep Worker." Explores codebases and executes goals without hand-holding. - Prometheus: The Strategic Planner. Interviews the user to refine requirements before execution starts. - Oracle: Specialized in architecture decisions and deep debugging.

Strengths¶

Multi-Model Orchestration: Routes tasks to the best model for the job (Claude for logic, Gemini for creativity, etc.).
Hashline (Hash-Anchored Edits): Edits lines by referencing content hashes, eliminating stale-line errors and whitespace issues.
Claude Code Compatibility: Support for Claude Code skills, hooks, and MCPs.
Transparent & Open: No "walled garden" lock-in; you control the providers and the tokens.

Limitations¶

Setup Complexity: While improved with init-deep, advanced multi-model configuration requires API key management for several providers.
Resource Usage: Running multiple specialized agents in parallel can consume more tokens than a single-agent approach.

When to use it¶

When you want a "Ubuntu" like experience for AI coding—stable, open, and powerful.
When surgical precision and high success rates for complex edits are more important than speed.

When not to use it¶

For trivial, single-file changes where a simple chat interface suffices.
If you prefer the simplicity of a single-provider, managed experience.

Licensing and cost¶

Open Source: Yes (SUL 1.0)
Cost: Free software; pay-per-token for whichever providers you use.
Self-hostable: Yes.

Getting started¶

Installation¶

You can install OmO via npm:

npm install -g oh-my-opencode

Initializing a Project¶

Run the deep initialization to set up hierarchical context for your agents:

/init-deep

CLI examples¶

# Start the "Ultrawork" loop (Plan -> Execute -> Verify)
ultrawork "Implement the authentication flow using NextAuth"

# Trigger a planning session with Prometheus
/start-work

# Self-referential loop until 100% done
/ulw-loop "Fix all linting errors and update dependencies"

# Run project diagnostics
bunx oh-my-opencode doctor

Sources / References¶

Contribution Metadata¶

Last reviewed: 2026-05-12
Confidence: high