OwO provides a set of plugins that can be used to supercharge OpenCode
You can use these plugins to create specialised AI agents that handle different aspects of development. Instead of one agent doing everything, you get a team of experts — each optimized for their domain. Or if you prefer, you can enhance the existing agents and configuration. Everything is customisable, so it's up to you!
Everything is opt-in. With OwO you have to configure things yourself, but the examples package is based on my own setup
- Background Tasks — Fire multiple agents in parallel
- System Prompt Injection - Add specific commands or skills directly in the system prompt, per agent
- Keyword Triggers —
ultrawork,deep research,explore codebase - Multi Agent Code review - For when just one agent isn't enough
- Github PR reviews - More control than the GH agent provides
- Various Tools - To provide more context and power to agents
- Example Skills - Based on Superpowers powerful skills for brainstorming, planning, and sub-agent powered builds
- Example Agents - Example agents, using zenox as a base: Explore, Librarian, Oracle, UI Planner. Pre-configured with Superpowers
OwO provides tools for tasks usually handled by MCP
- Context7 - For the latest documentation, handled as a single tool call. No looping around with ids
- Exa - Fast web and github search
- Jira - Read only access to tickets and dependencies
- Grep-App - Super fast github code search
- Coderabbit - Code reviews directly from the Coderabbit CLI
OwO takes the more lightweight sub-agent and multi-model approach from zenox. This lets you configure each agent to a model more tuned for the task.
For example, you can configure orchestration to have something like:
- Explore finds code fast — optimized for codebase search with a lightweight model (built into OpenCode, but you can use the
prompt-injectorto enhance it) - Librarian digs deep into docs — researches libraries, finds GitHub examples, citations included
- Oracle thinks strategically — architecture decisions, debugging, technical trade-offs
- UI Planner designs beautifully — CSS, animations, interfaces that don't look AI-generated
Or you could decide to merge Explore/Librarian into one. Maybe you have no UI work, so need a database expert instead. You can do as you please
You can set your main agent to automatically delegate to specialists when needed. Or — better yet — keep the build agent lean and create an orchestrator agent to coordinate!
See packages/example for complete example configurations.
Things are still under development. You'll need to run from source for now:
git clone git@github.com:rawtoast/OwO.git
bun installMake note of where you do this, as you'll need the location for plugin installation.
To enable agent delegation (where one agent coordinates others), you need:
-
The orchestration plugin — Add
@owo/orchestrationto your plugins -
An orchestrator agent — Either:
- Create a custom agent with orchestration instructions in its prompt (see
packages/example/agent/owO.mdfor an example) - Or use the
@owo/prompt-injectorplugin to inject orchestration context into existing agents likebuildorplan
- Create a custom agent with orchestration instructions in its prompt (see
Check out the example configs in packages/example:
owo.example.json— Full OwO config with keywords, prompts, orchestration, and tool settingsopencode.example.json— OpenCode config showing how to wire up plugins and configure multiple agents with different models and permissions
You don't need to call agents directly. The main agent (Build/Plan) automatically delegates:
You: "Where's the authentication logic?"
→ Explorer searches the codebase
You: "How does React Query handle caching?"
→ Librarian fetches official docs + real examples
You: "Should I use Redux or Zustand here?"
→ Oracle analyzes trade-offs for your codebase
You: "Make this dashboard look better"
→ UI Planner redesigns with proper aesthetics
- OpenCode — The CLI this extends
- oh-my-opencode — Inspiration for orchestration patterns
- zenox - Originally forked form this great setup