OpenCode

OpenCode supports arbitrary OpenAI-compatible providers via its config. Add base/mnt.ai once and select it per-session.

Provider config

Global config lives at ~/.config/opencode/opencode.json (or opencode.json in a repo for project scope):

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "basemnt": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "base/mnt.ai",
      "options": {
        "baseURL": "https://api.basemnt.ai/v1",
        "apiKey": "{env:SK_KEY}"
      },
      "models": {
        "z-ai/glm-5.2": {},
        "deepseek/deepseek-v4-flash-0731": {},
        "minimax/minimax-m3": {}
      },
      "default": true
    }
  }
}

"{env:SK_KEY}" resolves the key from your environment at startup — keep the secret out of the config file:

export SK_KEY=sk-your-key

Bash function alternative

If you prefer OpenCode’s built-in env handling, wrap the binary:

opencode() {
  OPENAI_BASE_URL=https://api.basemnt.ai/v1 \
  OPENAI_API_KEY="$SK_KEY" \
  command opencode "$@"
}

Model selection

Inside a session, the provider’s models appear by ID:

opencode run --model basemnt/z-ai/glm-5.2 "explain the router auth flow"

"default": true on the provider makes basemnt models the default set.

Verify

Run opencode and check the model picker shows basemnt/z-ai/glm-5.2. A 401 means SK_KEY isn’t set in the environment OpenCode inherited; a 400 means the model ID isn’t in the catalog.