
oh-my-opencode-slim, a lite agent orchestration plugin
review of oh-my-opencode slim
Hi this is my first post and I will talk about oh-my-opencode-slim. I was looking for a way to make a multi agent setup and found this plugin.
What is oh-my-opencode-slim?
oh-my-opencode-slim is an agent orchestration plugin for OpenCode made by alvinunreal. It includes a built-in team of specialized agents that can scout a codebase, look up fresh documentation, review architecture, handle UI work, and execute well-scoped implementation tasks under one orchestrator.
There are seven specialized agents and you can choose models for the agents considering intelligence and cost.
The agents
There is Orchestrator, Explorer, Oracle, Council, Librarian, Designer, and Fixer.
Orchestrator
Orchestrator will be the one who you will interact with. It will make plans and delegate tasks to spcialized agents.
You would want a strong model capable of orchestration.
Explorer
Explorer searches codebases.
A fast, low-cost model would be a good fit.
Oracle
Oracle is the Strategic advisor and debugger of last resort.
Choose your strongest model.
Librarian
Librarian handles research and documentation lookups. Speed and efficiency matters.
Designer
Implements UI/UX. Choose a model that is strong at UI/UX judgment, frontend implementation, and visual polish.
Fixer
Handels scoped implementation work. Choose a model that excecuts plans well
Best models?
The plugin does have presets but besides the OpenAI preset it is a bit outdated. Here are my custom presets that I have been trying out.
$30 Preset
{
"presets": {
"thirtydollars": {
"orchestrator": {
"model": "openai/gpt-5.6-terra",
"skills": ["*"],
"mcps": ["*", "!context7"]
},
"oracle": {
"model": "openai/gpt-5.6-sol",
"skills": ["simplify"],
"mcps": []
},
"explorer": {
"model": "openai/gpt-5.6-luna",
"skills": [],
"mcps": []
},
"librarian": {
"model": "opencode-go/mimo-v2.5",
"skills": [],
"mcps": ["context7", "gh_grep"]
},
"designer": {
"model": "opencode-go/muse-spark-1.3-contributor",
"skills": [],
"mcps": []
},
"fixer": {
"model": "opencode-go/glm-5.3-flash",
"skills": [],
"mcps": []
}
}
}
}
So a “strong brain + inexpensive workforce” setup. It would help reducing ChatGPT plus usage.
Also I changed Github Copilot pro that the current $30 preset uses to OpenCode Go.
You could also fit Astra with low reasoning for Oracle.
opencode-go
"opencode-go": {
"orchestrator": {
"model": "opencode-go/glm-5.3-flash",
"variant": "high",
"skills": ["*"],
"mcps": ["*", "!context7"]
},
"oracle": {
"model": "opencode-go/glm-5.3-flash",
"variant": "max",
"skills": ["simplify"],
"mcps": []
},
"explorer": {
"model": "opencode-go/mimo-v2.5",
"skills": [],
"mcps": []
},
"librarian": {
"model": "opencode-go/mimo-v2.5",
"skills": [],
"mcps": ["context7", "gh_grep"]
},
"designer": {
"model": "opencode-go/muse-spark-1.3-contributor",
"variant": "max",
"skills": [],
"mcps": []
},
"fixer": {
"model": "opencode-go/muse-spark-1.3-contributor",
"skills": [],
"mcps": []
}
}
This one is an OpenCode Go only config. This set up is GLM 5.3 flash for reasoning, Muse Spark 1.3 for implementing, Mimo v2.5 for searching.
Currently I think these 3 models have the best value for price in OpenCode Go plan.
Final Thoughts
I think it is a solid option for people who want to try agent orchestration like me.
Also watching agents talking to each other is kind of fun!