Skip to main content

OpenCode

OpenCode has Standard Compute built in as a provider.

1. Install and open OpenCode​

curl -fsSL https://opencode.ai/install | bash

Then run opencode inside your project.

2. Connect your API key​

Run /connect, choose Standard Compute, and paste your API key from the dashboard.

3. Select the model​

Choose Standard Compute when the model picker opens. That's it.

You can change which models smart routing uses at any time on the Models page.

Manual configuration​

If Standard Compute doesn't appear in /connect, update OpenCode or add the provider yourself. This example is for OpenCode V1:

~/.config/opencode/opencode.json
{
  "$schema": "https://opencode.ai/config.json",
  "model": "standardcompute/standardcompute",
  "provider": {
    "standardcompute": {
      "npm": "@openrouter/ai-sdk-provider",
      "name": "standardcompute",
      "options": {
        "baseURL": "https://api.stdcmpt.com/v1",
        "apiKey": "{env:STANDARDCOMPUTE_API_KEY}"
      },
      "models": {
        "standardcompute": {
          "name": "standardcompute",
          "provider": { "npm": "@openrouter/ai-sdk-provider" }
        }
      }
    }
  }
}

Then set the key in your shell:

export STANDARDCOMPUTE_API_KEY=YOUR_API_KEY

For V1, use the OpenRouter adapter at both provider and model level. It preserves the structured reasoning metadata needed when sending tool results back. If replacing an older generic-adapter configuration, update both provider.standardcompute.npm and any provider.standardcompute.models.standardcompute.provider.npm override. Check your existing opencode.json or opencode.jsonc in both the global configuration directory and the project: an explicit override can take precedence over the built-in catalog even after an OpenCode update. Preserve your API key and other settings, restart OpenCode, and start a fresh conversation.

The agent config generator produces the appropriate configuration for OpenCode V1 or V2.