Setting Up oh-my-opencode

Install and configure oh-my-opencode plugin with platform-appropriate model routing.

Prerequisites

1. OpenCode installed via home-manager (from llm-agents.nix)
2. Authentication configured:
   • ChatGPT Plus: opencode auth login → OpenAI → headless
   • Vertex AI (macOS): gcloud auth application-default login

Platform Subscription Summary

Verify Vertex AI Access (macOS only)

Before installing, confirm Vertex AI credentials work:

# Check env vars
echo "GOOGLE_CLOUD_PROJECT: $GOOGLE_CLOUD_PROJECT"
echo "VERTEX_LOCATION: $VERTEX_LOCATION"

# Test model access
opencode models | grep google-vertex
opencode run --model google-vertex/gemini-2.5-flash "Say hello"
opencode run --model google-vertex-anthropic/claude-sonnet-4@20250514 "Say hello"

Installation

macOS (Vertex AI + ChatGPT + Copilot)

npx oh-my-opencode install --no-tui \
  --claude=yes \
  --openai=yes \
  --gemini=yes \
  --copilot=yes

Devbox (Max 20x + ChatGPT)

npx oh-my-opencode install --no-tui \
  --claude=max20 \
  --openai=yes

Post-Install: Configure Vertex AI Models (macOS)

The installer configures anthropic/ and google/ model names, but macOS should route through Vertex AI for work billing.

Edit ~/.config/opencode/oh-my-opencode.json:

Replace all occurrences in both agents and categories sections.

Verification

# Check agent assignments
cat ~/.config/opencode/oh-my-opencode.json | jq '.agents.sisyphus, .agents.oracle'

# Test a Vertex AI model
opencode run --model google-vertex-anthropic/claude-opus-4-5@20251101 "Say 'Vertex AI working'"

Troubleshooting