Use opencode with an uncensored LLM API

opencode is a terminal coding agent. It treats any OpenAI-compatible server as a custom provider that you declare in opencode.json, and it runs the request through the AI SDK's compatible package.

Set it up

1. Sign in at wildwestapi.com/signup/ with the six-digit code sent to your email, open the dashboard and create an API key. It starts with sk-ww- and is shown in full once, so copy it somewhere safe.
2. Start opencode and run /connect, scroll to Other, and enter a provider id such as wildwest.
3. Paste your sk-ww- key when asked. opencode stores it in ~/.local/share/opencode/auth.json.
4. Create opencode.json in your project root, or ~/.config/opencode/opencode.json for all projects.
5. Add the provider block below, using the same id you gave /connect.
6. Run /models and choose Outlaw 1, or set the model key to wildwest/outlaw-1.

opencode.json

For a /v1/chat/completions endpoint, the docs say to use the @ai-sdk/openai-compatible package. Models are declared in a map keyed by id.

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "wildwest": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Wild West API",
      "options": {
        "baseURL": "https://wildwestapi.com/v1",
        "apiKey": "{env:WILDWEST_API_KEY}"
      },
      "models": {
        "outlaw-1": {
          "name": "Outlaw 1",
          "limit": { "context": 500000, "output": 8192 }
        }
      }
    }
  },
  "model": "wildwest/outlaw-1"
}

The {env:WILDWEST_API_KEY} form reads the variable at startup, so the key stays out of the file. Note that an unset variable becomes an empty string, not an error, so a 401 usually means the variable was not exported in the shell that launched opencode. Model keys have the form provider/model.

Choosing the npm package

The same docs say to use @ai-sdk/openai only for providers that serve /v1/responses. Wild West documents the Chat Completions and Messages paths, so keep @ai-sdk/openai-compatible.

Adding more models

Add one entry under models per id you want, for example glm-5.3-flash-outlaw as a cheaper choice for small edits. Set limit.context to the real window: 1M for most, 512K for qwen3.8-27b-outlaw. The output value is your own cap on reply length, not an API limit.

One limit to know

The provider id in /connect must match the key under provider in the config, or opencode will not find the credential. And the model picker shows only models you declared, so a new Wild West model will not appear until you add it.

FAQ

Related