oh-my-opencode-slim

# oh-my-opencode-slim **A lightweight, powerful agent orchestration plugin for OpenCode** *Transform your AI assistant into a manager capable of delegating complex tasks to specialized sub-agents, running searches in the background, and managing multi-step workflows with ease.*

Slimmed-down fork of oh-my-opencode - focused on core agent orchestration with low token consumption.

Antigravity subscription recommended. The pantheon is tuned for Antigravity's model routing. Other providers work, but you'll get the best experience with Antigravity.

---

⚡ Quick Navigation

• 🚀 Installation
  • For Humans
  • For LLM Agents
• 🏗️ Architecture & Flow
• 🏛️ Meet the Pantheon
  • Orchestrator
  • Explorer
  • Oracle
  • Librarian
  • Designer
  • Fixer
• 🧩 Skills
  • Available Skills
  • Default Skill Assignments
  • YAGNI Enforcement
  • Playwright Integration
  • Customizing Agent Skills
• 🛠️ Tools & Capabilities
  • Tmux Integration
  • Quota Tool
  • Background Tasks
  • LSP Tools
  • Code Search Tools
• 🔌 MCP Servers
• ⚙️ Configuration
  • Files You Edit
  • OpenCode Config
  • Plugin Config
• 🗑️ Uninstallation

---

Installation

For Humans

Run the interactive installer:

bunx oh-my-opencode-slim install

Or use non-interactive mode:

bunx oh-my-opencode-slim install --no-tui --antigravity=yes --openai=yes --tmux=no

After installation, authenticate with your providers:

opencode auth login
# Select your provider → Complete OAuth flow
# Repeat for each provider you enabled

Once authenticated, run opencode and ping all agents to verify all agents respond.

💡 Tip: Models are fully customizable. The installer sets sensible defaults, but you can assign any model to any agent. Edit ~/.config/opencode/oh-my-opencode-slim.json to override models, adjust reasoning effort, or disable agents entirely. See Configuration for details.

Alternative: Ask any coding agent

Paste this into Claude Code, AmpCode, Cursor, or any coding agent:

Install and configure by following the instructions here:
https://raw.githubusercontent.com/alvinunreal/oh-my-opencode-slim/refs/heads/master/README.md

For LLM Agents

Instructions for LLM Agents (click to expand)

If you're an LLM Agent helping set up oh-my-opencode-slim, follow these steps.

---

Step 1: Check OpenCode Installation

opencode --version

If not installed, direct the user to https://opencode.ai/docs first.

---

Step 2: Ask About Provider Access

Ask these questions one at a time, waiting for responses:

1. "Do you have an Antigravity subscription?" (Provides Claude + Gemini via google/ prefix)
2. "Do you have access to OpenAI API?" (Enables openai/ models)

Help the user understand the tradeoffs:

• Antigravity is the recommended path for Claude + Gemini routing.
• OpenAI is optional; it enables openai/ models.
• If the user has no providers, the plugin still works using OpenCode Zen free models. They can switch to paid providers later by editing ~/.config/opencode/oh-my-opencode-slim.json.

---

Step 3: Run the Installer

Based on answers, run:

bunx oh-my-opencode-slim install --no-tui --antigravity=<yes|no> --openai=<yes|no>

Examples:

# Antigravity + OpenAI
bunx oh-my-opencode-slim install --no-tui --antigravity=yes --openai=yes --tmux=no

# OpenAI only
bunx oh-my-opencode-slim install --no-tui --antigravity=no --openai=yes --tmux=no

# No providers (Zen free models only)
bunx oh-my-opencode-slim install --no-tui --antigravity=no --openai=no --tmux=no

The installer automatically:

• Adds the plugin to ~/.config/opencode/opencode.json
• Adds opencode-antigravity-auth plugin (if Antigravity enabled)
• Configures Google provider with model definitions
• Generates agent model mappings in ~/.config/opencode/oh-my-opencode-slim.json

Crucial Advice for the User:

• They can easily assign different models to different agents by editing ~/.config/opencode/oh-my-opencode-slim.json.
• If they add a new provider later, they just need to update this file.
• Read generated ~/.config/opencode/oh-my-opencode-slim.json file and report the model configuration.

---

Step 4: Authentication

After installation, guide the user:

For Antigravity/OpenAI:

opencode auth login
# Select: Google → OAuth with Google (Antigravity)
# Repeat for all other providers

---

Troubleshooting

If the installer fails, check the expected config format:

bunx oh-my-opencode-slim install --help

Then manually create the config files at:

• ~/.config/opencode/oh-my-opencode-slim.json

---

🏗️ Architecture & Flow

The plugin follows a "Hub and Spoke" model:

1. The Orchestrator (Hub): The main entry point for user requests. It analyzes the task and decides which specialized agents to call.
2. Specialized Agents (Spokes): Domain-specific experts (e.g., UI/UX, Documentation, Architecture) that handle narrow tasks with high precision.
3. Background Manager: A robust engine that allows the Orchestrator to "fire and forget" tasks (like deep codebase searches or documentation research) while continuing to work on other parts of the problem.

🏛️ The Flow of a Request

1. User Prompt: "Refactor the auth logic and update the docs."
2. Orchestrator: Creates a TODO list.
3. Delegation:
   • Launches an @explorer background task to find all auth-related files.
   • Launches a @librarian task to check the latest documentation for the auth library used.
4. Integration: Once background results are ready, the Orchestrator delegates to @fixer to perform the refactor efficiently.

---

Primary vs Subagents

Primary agents appear in the OpenCode UI selector, while subagents are hidden from the UI and intended for delegation through background_task.

Primary agents (UI):

• OpenCode built-ins: Build, Plan (disabled by the installer by default)
• Plugin primary: orchestrator

Subagents (background_task): explorer, librarian, oracle, designer, fixer

The background_task tool lists only subagents. If the UI list looks stale after changes, restart OpenCode.

---

Meet the Pantheon

Orchestrator

The Orchestrator was born when the first codebase collapsed under its own complexity. Neither god nor mortal would claim responsibility - so The Orchestrator emerged from the void, forging order from chaos. They don't merely command armies; they fight alongside them. Every line of code passes through their hands before they decide which lesser deity deserves a piece of the puzzle.

Role: Supreme executor, delegator, and overseer
Model: google/claude-opus-4-5-thinking
Prompt: src/agents/orchestrator.ts

Write and execute code, orchestrate multi-agent workflows, parse the unspoken from the spoken, summon specialists mid-battle. Shape reality directly - and assign realms to others when the universe grows too vast.

---

Explorer

The Explorer moves through codebases like wind through trees - swift, silent, everywhere at once. When The Orchestrator whispers "find me the auth module," The Explorer has already returned with forty file paths and a map. They were born from the first grep command, evolved beyond it, and now see patterns mortals miss.

Role: Codebase reconnaissance
Model: google/gemini-3-flash
Prompt: src/agents/explorer.ts

Regex search, AST pattern matching, file discovery, parallel exploration. Read-only: they chart the territory; others conquer it.

---

Oracle

The Oracle does not code - they know. When bugs defy logic and architectures crumble, The Oracle gazes into the abyss of your codebase and speaks truth. They've seen a thousand systems rise and fall. They'll tell you which path leads to ruin, and which to production.

Role: Strategic advisor and debugger of last resort
Model: openai/gpt-5.2-codex
Prompt: src/agents/oracle.ts

Root cause analysis, architecture review, debugging guidance, tradeoff analysis. Read-only: Oracles advise; they don't intervene.

---

Librarian

The Librarian guards a library with no walls - every GitHub repo, every npm package, every StackOverflow answer ever written. Ask them "how does React handle concurrent rendering?" and they'll return with official docs, real-world examples, and a warning about the footgun you're about to step on.

Role: External knowledge retrieval
Model: google/gemini-3-flash
Prompt: src/agents/librarian.ts

Documentation lookup, GitHub code search, library research, best practice retrieval. Read-only: they fetch wisdom; implementation is for others.

---

Designer

The Designer believes code should be beautiful - and so should everything it renders. Born from the frustration of a thousand ugly MVPs, they wield CSS like a brush and components like clay. Hand them a feature request; receive a masterpiece. They don't do "good enough."

Role: UI/UX implementation and visual excellence
Model: google/gemini-3-flash
Prompt: src/agents/designer.ts

Modern responsive design, CSS/Tailwind mastery, micro-animations, component architecture. Visual excellence over code perfection - beauty is the priority.

---

Fixer

The Fixer is the hands that build what others envision. While The Orchestrator plans and The Oracle advises, The Fixer executes. They receive complete context from research agents and clear task specifications, then implement with surgical precision. Fast, efficient, and focused - they don't think about what to build, they just build it.

Role: Fast implementation specialist
Model: google/gemini-3-flash
Prompt: src/agents/fixer.ts

Code implementation, refactoring, testing, verification. Execute the plan - no research, no delegation, no planning.

---

Tools & Capabilities

Tmux Integration

⚠️ Known Issue: When the server port is enabled, only one OpenCode instance can be opened at a time. We're tracking this in issue #15, and there's an upstream PR to OpenCode: opencode#9099.

Watch your agents work in real-time. When the Orchestrator launches sub-agents or initiates background tasks, new tmux panes automatically spawn showing each agent's live progress. No more waiting in the dark.

Why This Matters

Without Tmux Integration	With Tmux Integration
Fire off a background task, wait anxiously	See the agent thinking, searching, coding
"Is it stuck or just slow?"	Watch tool calls happen in real-time
Results appear out of nowhere	Follow the journey from question to answer
Debug by guessing	Debug by observation

What You Get

• Live Visibility: Each sub-agent gets its own pane showing real-time output
• Auto-Layout: Tmux automatically arranges panes using your preferred layout
• Auto-Cleanup: Panes close when agents finish, layout rebalances
• Zero Overhead: Works with OpenCode's built-in task tool AND our background_task tool

Quick Setup

1. Enable the OpenCode HTTP server in opencode.json (see OpenCode Config).
2. Enable tmux integration in oh-my-opencode-slim.json (see Plugin Config).
3. Run OpenCode inside tmux: bash tmux opencode

Layout Options

Layout	Description
main-vertical	Your session on the left (60%), agents stacked on the right
main-horizontal	Your session on top (60%), agents stacked below
tiled	All panes in equal-sized grid
even-horizontal	All panes side by side
even-vertical	All panes stacked vertically

See the Option Reference for detailed configuration.

---

Quota Tool

For Antigravity users. You can trigger this at any time by asking the agent to "check my quota" or "show status."

Tool	Description
antigravity_quota	Check API quota for all Antigravity accounts (compact view with progress bars)

---

Background Tasks

The plugin provides tools to manage asynchronous work:

Tool	Description
background_task	Launch an agent in a new session (sync=true blocks, sync=false runs in background)
background_output	Fetch the result of a background task by ID
background_cancel	Abort running tasks

---

LSP Tools

Language Server Protocol integration for code intelligence:

Tool	Description
lsp_goto_definition	Jump to symbol definition
lsp_find_references	Find all usages of a symbol across the workspace
lsp_diagnostics	Get errors/warnings from the language server
lsp_rename	Rename a symbol across all files

---

Code Search Tools

Fast code search and refactoring:

Tool	Description
grep	Fast content search using ripgrep
ast_grep_search	AST-aware code pattern matching (25 languages)
ast_grep_replace	AST-aware code refactoring with dry-run support

---

🧩 Skills

Skills are specialized capabilities that agents can use. Each agent has a default set of skills, which you can override in the agent config.

Available Skills

Skill	Description
yagni-enforcement	Code complexity analysis and YAGNI enforcement
playwright	Browser automation via Playwright MCP

Default Skill Assignments

Agent	Default Skills
orchestrator	* (all skills)
designer	playwright
oracle	none
librarian	none
explorer	none
fixer	none

YAGNI Enforcement

The Minimalist's sacred truth: every line of code is a liability.

Use after major refactors or before finalizing PRs. Identifies unnecessary complexity, challenges premature abstractions, estimates LOC reduction, and enforces minimalism.

Playwright Integration

Browser automation for visual verification and testing.

• Browser Automation: Full Playwright capabilities (browsing, clicking, typing, scraping).
• Screenshots: Capture visual state of any web page.
• Sandboxed Output: Screenshots saved to session subdirectory (check tool output for path).

Customizing Agent Skills

Override skills per-agent in your Plugin Config:

{
  "agents": {
    "orchestrator": {
      "skills": ["*"]
    },
    "designer": {
      "skills": ["playwright"]
    }
  }
}

---

MCP Servers

Built-in Model Context Protocol servers (enabled by default):

MCP	Purpose	URL
websearch	Real-time web search via Exa AI	https://mcp.exa.ai/mcp
context7	Official library documentation	https://mcp.context7.com/mcp
grep_app	GitHub code search via grep.app	https://mcp.grep.app

Disabling MCPs

You can disable specific MCP servers by adding them to the disabled_mcps array in your Plugin Config.

---

Configuration

Files You Edit

File	Purpose
~/.config/opencode/opencode.json	OpenCode core settings (server port for tmux)
~/.config/opencode/oh-my-opencode-slim.json	Plugin settings (agents, tmux, MCPs)
.opencode/oh-my-opencode-slim.json	Project-local plugin overrides (optional)

Platform paths: On Windows, user config may also be at %APPDATA%\opencode\.

---

Plugin Config (oh-my-opencode-slim.json)

The installer generates this file based on your providers. You can manually customize it to mix and match models.

Example: Antigravity + OpenAI (Recommended)

{
  "agents": {
    "orchestrator": { "model": "google/claude-opus-4-5-thinking", "skills": ["*"] },
    "oracle": { "model": "openai/gpt-5.2-codex", "skills": [] },
    "librarian": { "model": "google/gemini-3-flash", "skills": [] },
    "explorer": { "model": "google/gemini-3-flash", "skills": [] },
    "designer": { "model": "google/gemini-3-flash", "skills": ["playwright"] },
    "fixer": { "model": "google/gemini-3-flash", "skills": [] }
  },
  "tmux": {
    "enabled": true,
    "layout": "main-vertical",
    "main_pane_size": 60
  }
}

Example: Antigravity Only

{
  "agents": {
    "orchestrator": { "model": "google/claude-opus-4-5-thinking", "skills": ["*"] },
    "oracle": { "model": "google/claude-opus-4-5-thinking", "skills": [] },
    "librarian": { "model": "google/gemini-3-flash", "skills": [] },
    "explorer": { "model": "google/gemini-3-flash", "skills": [] },
    "designer": { "model": "google/gemini-3-flash", "skills": ["playwright"] },
    "fixer": { "model": "google/gemini-3-flash", "skills": [] }
  }
}

Example: OpenAI Only

{
  "agents": {
    "orchestrator": { "model": "openai/gpt-5.2-codex", "skills": ["*"] },
    "oracle": { "model": "openai/gpt-5.2-codex", "skills": [] },
    "librarian": { "model": "openai/gpt-5.1-codex-mini", "skills": [] },
    "explorer": { "model": "openai/gpt-5.1-codex-mini", "skills": [] },
    "designer": { "model": "openai/gpt-5.1-codex-mini", "skills": ["playwright"] },
    "fixer": { "model": "openai/gpt-5.1-codex-mini", "skills": [] }
  }
}

Option Reference

Option	Type	Default	Description
tmux.enabled	boolean	false	Enable tmux pane spawning for sub-agents
tmux.layout	string	"main-vertical"	Layout preset: main-vertical, main-horizontal, tiled, even-horizontal, even-vertical
tmux.main_pane_size	number	60	Main pane size as percentage (20-80)
disabled_agents	string[]	[]	Agent IDs to disable (e.g., "explorer")
disabled_mcps	string[]	[]	MCP server IDs to disable (e.g., "websearch")
agents.<name>.model	string	—	Override the LLM for a specific agent
agents.<name>.variant	string	—	Reasoning effort: "low", "medium", "high"
agents.<name>.skills	string[]	—	Skills this agent can use ("*" = all)
agents.<name>.temperature	number	—	Temperature for this agent (0.0 to 2.0)
agents.<name>.prompt	string	—	Base prompt override for this agent
agents.<name>.prompt_append	string	—	Text to append to the base prompt
agents.<name>.disable	boolean	—	Disable this specific agent

---

Uninstallation

1. Remove the plugin from your OpenCode config:

Edit ~/.config/opencode/opencode.json and remove "oh-my-opencode-slim" from the plugin array.

1. Remove configuration files (optional): bash rm -f ~/.config/opencode/oh-my-opencode-slim.json rm -f .opencode/oh-my-opencode-slim.json

---

Credits

This is a slimmed-down fork of oh-my-opencode by @code-yeongyu.

---

License

MIT