Configure OpenCode

Point the OpenCode CLI at your local Sombra proxy.

1. Add the provider

OpenCode is configured via opencode.json in the project root. This file is already shipped in the repo — it registers the local LiteLLM proxy as a provider and selects its PII-protected model by default:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "litellm-proxy": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Sombra LiteLLM Proxy",
      "options": {
        "baseURL": "http://localhost:4000/v1",
        "apiKey": "{env:LITELLM_MASTER_KEY}"
      },
      "models": {
        "opencode-zen/deepseek-v4-flash": {
          "name": "Sombra - DeepSeek v4 Flash (PII Protected)"
        }
      }
    }
  },
  "model": "litellm-proxy/opencode-zen/deepseek-v4-flash"
}

2. Export the master key

OpenCode does not auto-load .env, so the proxy's master key must be in your shell before starting OpenCode:

set -a; source .env; set +a

3. Start OpenCode

opencode run "Consult my order in the name of João, CPF 123.456.789-00, living in Curitiba, CEP 80000-000."

Every request to the model now passes through Sombra's masking guardrail before leaving your machine.

Next

Your setup is complete. Verify it works with Use it.