2026-06-03 – Install & Configure oh‑my‑openagent (OpenCode)¶

Context – This guide assumes you already have an OVHcloud AI Endpoints subscription configured in OpenCode (the "OVH" article in content/done/20260603‑opencode‑ovh‑oh‑my‑opencode.md). We therefore skip all subscription questions and directly run the Ultimate installer.

Prerequisites¶

OpenCode version ≥ 1.0.150 installed and on your $PATH.
Bun (the JS runtime) – required for the installer (bunx).
Network connectivity to https://github.com/code-yeongyu/oh‑my‑openagent.

Verify OpenCode:

opencode --version

If the command fails, install OpenCode first (see the OVH article).

Step 0 – Prepare a non‑interactive install¶

Because the OVH subscription is already active we can tell the installer to skip all subscription prompts:

export OMO_SKIP_SUBSCRIPTIONS=1   # tell the installer to skip the subscription interview

The installer also respects --skip-auth to bypass provider OAuth prompts.

Step 1 – Run the Ultimate installer¶

bunx oh‑my‑openagent install \
  --no‑tui \
  --platform=opencode \
  --skip‑auth \
  --claude=no \
  --gemini=no \
  --copilot=no

The --claude/--gemini/--copilot flags are required by the installer even when OVH is our only provider. We set them to no.

What the command does:

Registers the plugin "oh‑my‑openagent" in opencode.json.
Writes a default model config (all agents → opencode/gpt-5-nano) — the installer doesn't auto-detect OVH.
We will replace that default with the actual OVH models in Step 3.

Step 2 – Verify the plugin is loaded¶

opencode --version   # should still print the same version
bunx oh‑my‑openagent doctor   # health‑check, should report no errors

You should see the plugin array in ~/.config/opencode/opencode.json contain:

"plugin": ["oh‑my‑openagent"]

Now confirm the OVH models are actually available:

opencode models | grep ovhcloud

You should see all the OVH models listed (qwen3.6-27b, gpt-oss-120b, qwen3.5-9b, qwen3.5-397b-a17b, etc.). If nothing shows up, run /connect inside OpenCode and re-authenticate OVH first.

Step 3 – Override the default model config with OVH recommendations¶

Replace the contents of ~/.config/opencode/oh‑my‑openagent.json (created by the installer) with the full OVH config below. This maps every category and agent to the OVH models, and enables Team Mode with 4 parallel agents:

{
  "$schema": "https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/dev/assets/oh-my-opencode.schema.json",
  "team_mode": {
    "enabled": true,
    "max_parallel_members": 4
  },
  "categories": {
    "visual-engineering": {
      "model": "ovhcloud/qwen3.6-27b",
      "fallback_models": [
        "ovhcloud/qwen2.5-vl-72b-instruct",
        "ovhcloud/mistral-small-3.2-24b-instruct-2506"
      ]
    },
    "deep": {
      "model": "ovhcloud/gpt-oss-120b",
      "fallback_models": [
        "ovhcloud/qwen3.5-397b-a17b",
        "ovhcloud/qwen3-32b"
      ]
    },
    "quick": {
      "model": "ovhcloud/qwen3.5-9b",
      "fallback_models": [
        "ovhcloud/gpt-oss-20b",
        "ovhcloud/mistral-7b-instruct-v0.3"
      ],
      "temperature": 0.3
    },
    "ultrabrain": {
      "model": "ovhcloud/qwen3.5-397b-a17b",
      "fallback_models": [
        "ovhcloud/gpt-oss-120b",
        "ovhcloud/meta-llama-3_3-70b-instruct"
      ],
      "temperature": 0.2
    },
    "artistry": {
      "model": "ovhcloud/qwen3.6-27b",
      "fallback_models": ["ovhcloud/gpt-oss-120b"]
    },
    "unspecified-low": {
      "model": "ovhcloud/qwen3.5-9b",
      "fallback_models": ["ovhcloud/gpt-oss-20b"]
    },
    "unspecified-high": {
      "model": "ovhcloud/gpt-oss-120b",
      "fallback_models": ["ovhcloud/qwen3.6-27b"]
    },
    "writing": {
      "model": "ovhcloud/qwen3.6-27b",
      "fallback_models": ["ovhcloud/qwen3-32b"]
    }
  },
  "agents": {
    "sisyphus": {
      "model": "ovhcloud/gpt-oss-120b",
      "fallback_models": ["ovhcloud/qwen3.6-27b", "ovhcloud/qwen3.5-397b-a17b"]
    },
    "hephaestus": {
      "model": "ovhcloud/gpt-oss-120b",
      "fallback_models": ["ovhcloud/qwen3.5-397b-a17b"]
    },
    "oracle": {
      "model": "ovhcloud/qwen3.5-397b-a17b",
      "fallback_models": ["ovhcloud/gpt-oss-120b"]
    },
    "explore": {
      "model": "ovhcloud/qwen3.5-9b",
      "fallback_models": ["ovhcloud/gpt-oss-20b"]
    },
    "librarian": {
      "model": "ovhcloud/qwen3.5-9b",
      "fallback_models": ["ovhcloud/gpt-oss-20b"]
    },
    "multimodal-looker": {
      "model": "ovhcloud/qwen3.6-27b",
      "fallback_models": ["ovhcloud/mistral-small-3.2-24b-instruct-2506"]
    },
    "prometheus": {
      "model": "ovhcloud/gpt-oss-120b",
      "fallback_models": ["ovhcloud/qwen3.5-397b-a17b"]
    },
    "metis": {
      "model": "ovhcloud/qwen3.6-27b",
      "fallback_models": ["ovhcloud/gpt-oss-120b"]
    },
    "momus": {
      "model": "ovhcloud/qwen3.5-397b-a17b",
      "fallback_models": ["ovhcloud/gpt-oss-120b"]
    },
    "atlas": {
      "model": "ovhcloud/gpt-oss-120b",
      "fallback_models": ["ovhcloud/qwen3.6-27b"]
    },
    "sisyphus-junior": {
      "model": "ovhcloud/gpt-oss-120b",
      "fallback_models": ["ovhcloud/qwen3.6-27b"]
    }
  }
}

Important: After writing the config, restart OpenCode (exit and start a new session) so the new model mappings and Team Mode take effect.

Step 4 – Confirm model availability¶

# List the OVH models that the OpenCode provider currently knows about
/opencode models list | grep ovhcloud

You should see at least the four primary models (qwen3.6‑27b, gpt‑oss‑120b, qwen3.5‑9b, qwen3.5‑397b‑a17b). If a model is missing, run:

opencode models refresh   # forces a fresh catalog download from OVH

Step 5 – Quick functional test¶

Run the built‑in doctor again (already done) and then a simple category test:

# Visual‑engineering test (frontend UI)
echo "Create a simple React button component" | opencode

# Quick test (single‑file edit)
echo "Fix the typo in README.md line 42" | opencode

The output should indicate the agent selected the appropriate category and used the corresponding model.

Step 6 – Team Mode (now enabled)¶

Team Mode is oh-my-openagent's multi-agent orchestration layer. When enabled (it is, in the config above), the lead agent can spawn up to max_parallel_members specialist agents running in parallel, communicating via dedicated team_* tools (team_create, team_send_message, team_task_create, etc.).

Two skills ride on Team Mode out of the box:

hyperplan — 5 hostile critics tear apart a plan from orthogonal angles before any code is written.
security-research — 3 vulnerability hunters + 2 PoC engineers audit your codebase in parallel.

Adjust max_parallel_members (1-8) in the config if needed. The tmux_visualization option (disabled by default) shows each member's workspace in a tmux grid.

Summary¶

Skip subscriptions – use --no‑tui --skip‑auth (and OMO_SKIP_SUBSCRIPTIONS=1).
Use OVH‑recommended models – set them in ~/.config/opencode/oh‑my‑openagent.json as shown.
Team Mode enabled by default with 4 parallel agents.
Restart OpenCode after any config change.
Verify with bunx oh‑my‑openagent doctor and a couple of quick prompts.

You are now ready to use oh‑my‑openagent with the same OVH model stack you already have configured for OpenCode.

Backlinks¶

The following pages link to this page:

Git AI Integration with OpenCode

2026-06-03 – Install & Configure oh‑my‑openagent (OpenCode)¶

Prerequisites¶

Step 0 – Prepare a non‑interactive install¶

Step 1 – Run the Ultimate installer¶

Step 2 – Verify the plugin is loaded¶

Step 3 – Override the default model config with OVH recommendations¶

Step 4 – Confirm model availability¶

Step 5 – Quick functional test¶

Step 6 – Team Mode (now enabled)¶