opencode-config

To configure OpenCode CLI providers and models, edit opencode.json at the project, global, or auth level to set your default model and provider. The configuration merges these locations using a prioritized project > global > auth hierarchy to determine final settings deterministically.

OpenCode CLI model selection reverts because configurations are scattered across project, global, and auth locations with a strict priority hierarchy. You need to audit and enforce the prioritized config model at the correct level to ensure deterministic model selection resolves the conflict.

Yes, you can use a custom OpenAI-compatible provider with OpenCode CLI by defining the provider options with a specific baseURL and setting mode-specific model mappings in your configuration file. This allows you to manage multiple OpenCode providers or environments.

The best way to troubleshoot OpenCode CLI configuration issues is to use the built-in troubleshooting workflows to audit your provider, model, and baseURL settings. After editing opencode.json, use the /models command to test changes and restart Claude to reload the config.

Yes, you need to restart Claude to apply new OpenCode CLI baseURL settings so the application can reload the configuration. If issues arise after modifying the provider options in opencode.json, consulting the documentation and restarting ensures the changes take effect.