OpenCode
Details for running the OpenCode harness inside Conductor
OpenCode is an open source AI coding harness. OpenCode supports any LLM provider (like OpenRouter, Baseten, Cerebras, Vercel AI Gateway, and more) as well as the latest open-source and frontier models.
OpenCode quick facts
| Topic | OpenCode |
|---|---|
| Availability | Bundled in Conductor. You can use your system OpenCode instead. |
| Setup | Settings -> Harnesses -> OpenCode for provider keys, visible models, and the OpenCode executable path. |
| Models | Open- and closed-source models returned by your OpenCode providers. |
| Billing | The selected OpenCode provider, local model, ChatGPT, or GitHub Copilot. |
| Controls | Skills are supported. Plan Mode, Fast Mode, goals, and checkpoints are not supported today. |
Setup notes
Conductor can use the bundled OpenCode executable or a configured system OpenCode executable.
Settings -> Harnesses -> OpenCode shows provider status, provider API keys, visible models, and the custom OpenCode executable path.
Conductor can use provider keys from:
- OpenCode provider API keys saved in Conductor, including keys for providers such as OpenRouter, Baseten, Cerebras, and Vercel AI Gateway
- local environment variables detected from your shell or Conductor environment settings
- OpenCode's own provider configuration
OpenCode itself supports /connect and opencode.json provider configuration. The linked OpenCode docs cover providers that are not already configured in Conductor.
For a setup walkthrough, see Configure model providers: OpenCode.
Model access
Conductor asks OpenCode for the models available to the configured setup. The list depends on the OpenCode executable, provider credentials, OpenCode config, and the models returned by OpenCode.
OpenCode model IDs are provider-qualified, such as anthropic/claude-sonnet-4-5 or openai/gpt-5.1. Conductor labels those as provider/model choices in the model picker.