Skip to content

Oh My OpenAgent (OmO)

Oh My OpenAgent (abbreviated as OmO, formerly oh-my-opencode) is a powerful agent-orchestration harness and plugin suite designed specifically for the OpenCode CLI. It elevates the standard OpenCode terminal experience by enabling multi-agent workflows, intelligent model routing, and advanced persistent memory.

When integrated with our Swiss-hosted LLMaaS Gateway, OmO allows you to coordinate complex software engineering projects locally while keeping all data, codebases, and AI reasoning steps fully within Switzerland.

Key Features

  • Multi-Agent Orchestration (Agent Teams): Coordinated workflows involving multiple specialized sub-agents:
    • Oracle: Specializes in project-wide planning, architecture analysis, and design decisions.
    • Hephaestus: Executes precise code modifications and file-system edits.
    • Sisyphus: Automates background tasks, runs compilers, linting tools, and testing suites.
  • Intelligent Model Routing: Automatically routes specific tasks to the most suitable model in your configuration. For instance, complex architectural plans are routed to reasoning models like ew/inference-deepseekr1-670b, whereas routine code edits are routed to faster coding models like ew/minimax27 or ew/inference-gemma4-31b.
  • Model Context Protocol (MCP) Support: Provides extended tool integrations and handles background context sharing with external servers.
  • Persistent Session State: Advanced "context forking" that lets you pause, branch out, or rewind your development sessions without losing state.

Installation

Oh My OpenAgent requires the Bun runtime and can be installed directly into your OpenCode setup via:

dpkg -l unzip || sudo apt install unzip
dpkg -l curl || sudo apt install curl
which bunx || (curl -fsSL https://bun.com/install | bash ; exec $SHELL)
which bunx && echo yeah - bunx is installed

Now we can proceed with the installation of oh-my-openagent. Additionally, install the dcp (Dynamic Context Pruning) plugin:

opencode plugin oh-my-openagent@3.14.0
opencode plugin @tarquinen/opencode-dcp@latest --global

Integration with LLMaaS Gateway

Once installed, Oh My OpenAgent utilizes the providers defined in your opencode.json configuration. It leverages the model routing capabilities by mapping task profiles to different LLMaaS models.

First, create a backup of your default oh-my-openagent configuration (only works if a configuration already exists):

cp ~/.config/opencode/oh-my-openagent.json{,.org}

Then edit the configuration:

cat <<'EOF' > ~/.config/opencode/oh-my-openagent.json
{
  "$schema": "https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/dev/assets/oh-my-opencode.schema.json",
  "agents": {
    "sisyphus": {
      "model": "ew/ew/qwen36-35b-a3b"
    },
    "hephaestus": {
      "model": "ew/ew/minimax27"
    },
    "oracle": {
      "model": "ew/ew/inference-deepseekr1-670b"
    },
    "explore": {
      "model": "ew/ew/inference-llama4-maverick"
    },
    "multimodal-looker": {
      "model": "ew/ew/inference-kimi-k2"
    },
    "prometheus": {
      "model": "ew/ew/inference-deepseekr1-670b"
    },
    "metis": {
      "model": "ew/ew/inference-deepseek-v32"
    },
    "momus": {
      "model": "ew/ew/inference-gpt-oss-120b"
    },
    "atlas": {
      "model": "ew/ew/inference-llama4-maverick"
    },
    "sisyphus-junior": {
      "model": "ew/ew/qwen36-35b-a3b"
    }
  },
  "categories": {
    "visual-engineering": {
      "model": "ew/ew/qwen36-35b-a3b"
    },
    "ultrabrain": {
      "model": "ew/ew/inference-deepseekr1-670b"
    },
    "deep": {
      "model": "ew/ew/inference-deepseek-v32"
    },
    "artistry": {
      "model": "ew/ew/inference-gemma4-31b"
    },
    "quick": {
      "model": "ew/ew/qwen36-35b-a3b"
    },
    "unspecified-low": {
      "model": "ew/ew/qwen36-35b-a3b"
    },
    "unspecified-high": {
      "model": "ew/ew/inference-deepseek-v32"
    },
    "writing": {
      "model": "ew/ew/inference-gpt-oss-120b"
    }
  }
}
EOF