OpenCode

Before you start#

Install OpenCode and obtain your plan API key, the provider's base URL, and the exact model name. The following example requires an OpenAI-compatible Chat Completions endpoint. Compatibility for this independent service has not yet been verified.

Add your key#

Check your installation first:

opencode --version

If it is not installed, use the OpenCode download page. Update an older client before troubleshooting provider settings.

Start OpenCode in your project and enter /connect. Choose Other, enter coding-plan as the provider ID, and enter your API key. OpenCode stores this credential locally; do not put the key in your project configuration.

Add the provider#

Create or update opencode.json in your project. Replace the URL and model identifier with the actual values supplied by the service. If the file already exists, merge these fields into the existing JSON instead of overwriting it.

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "coding-plan": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Coding Plan",
      "options": {
        "baseURL": "https://your-service-base-url"
      },
      "models": {
        "your-model-name": { "name": "Your model" }
      }
    }
  }
}

The provider ID in the file must match the ID used with /connect. The base URL must include any version prefix required by the service.

Select and test the model#

Restart OpenCode if needed, enter /models, and select your configured model. Send a small test request. If it fails, verify the model ID, URL, key, API format, and remaining allowance.

Common configuration mistakes#

To switch back, select your previous provider with /models. Remove only the custom provider entry if you no longer need it, preserving the rest of your configuration.

Official reference#

See OpenCode custom providers and DeepSeek's OpenCode guide. The built-in DeepSeek provider uses official DeepSeek credentials and may not accept credentials from this service.