OpenCode

OpenCode is an open-source AI coding assistant. Point it at your mycellm node for distributed inference.

Setup

# Set mycellm as the LLM backend
export OPENAI_BASE_URL=http://localhost:8420/v1
export OPENAI_API_KEY=your-mycellm-key  # optional
export OPENAI_MODEL=auto

# Start OpenCode
opencode

Configuration file

In your OpenCode config (~/.config/opencode/config.json or project .opencode.json):

{
  "provider": "openai",
  "model": "auto",
  "apiBase": "http://localhost:8420/v1",
  "apiKey": "your-mycellm-key"
}

Using with a remote mycellm node

If your mycellm node is on another machine (e.g., a GPU server):

export OPENAI_BASE_URL=http://gpu-server:8420/v1
opencode

Using with the public network

No mycellm node needed — use the public gateway directly:

export OPENAI_BASE_URL=https://api.mycellm.dev/v1/public
opencode

Note

The public gateway is rate-limited (5,000 tokens/day). For higher limits, run your own node — contributors who seed compute earn credits for inference on bigger models.

Tips

• Use auto as the model name — mycellm routes to the best available
• If you have multiple models loaded, specify one by name for consistency
• The mycellm dashboard (:8420) shows which models are available