OpenCode configuration, ready to copy.
Connect an OpenAI-compatible Chat Completions endpoint to OpenCode. Choose your version, add your provider and model, then download the configuration. Works with Standard Compute or your own compatible provider.
Check your version with opencode --version. V1 and V2 use different configuration fields.
Enter the variable name only. Set its value locally before starting OpenCode. This tool does not need your API key.
{
"$schema": "https://opencode.ai/config.json",
"model": "standardcompute/standardcompute",
"provider": {
"standardcompute": {
"npm": "@ai-sdk/openai-compatible",
"name": "standardcompute",
"options": {
"baseURL": "https://api.stdcmpt.com/v1",
"apiKey": "{env:STANDARDCOMPUTE_API_KEY}"
},
"models": {
"standardcompute": {
"name": "standardcompute"
}
}
}
}
}
Generated in your browser. No signup required.
Use your generated configuration
- Set the API key environment variable locally using your shell or secret manager. Start OpenCode from the environment that contains it.
- Merge the generated fields into your project's
opencode.json, or your global~/.config/opencode/opencode.json. Keep existing provider, permission and other settings. - Restart OpenCode and select the provider/model in
/models. Run a small task to check authentication, tool calls and streaming before a large job.
This generator prepares configuration; it does not connect to your provider or validate model access. Use the provider's exact model ID and base URL. Responses-only or Anthropic Messages endpoints need a different adapter.
For a custom model, check its context limit, image support and other capabilities in the provider's documentation before larger tasks. The minimal export does not supply those model-specific settings. V1's public schema may flag a custom default model that is absent from its built-in catalog.
Why are there two formats?
V1 uses provider, npm and options. V2 uses providers, package and settings. Choose the format for your installed version. The V2 example follows the V2 documentation; the shared public JSON schema currently describes V1, so the V2 download omits that schema reference.
Configuration references checked September 5, 2026. V1 providers · V1 environment variables · V2 providers · V2 configuration.
Compare what a completed task actually costs
Try the same task and repository revision with each setup. Record retries, token categories, elapsed time and whether the result passed your acceptance check. Keep subscription fees separate from token-price estimates.
Want a predictable monthly API bill?
Standard Compute starts at $19/month with a $20 monthly compute budget. Usage stops when the budget is exhausted; there are no automatic overage charges. Routing and your workload determine how far the budget goes.