Use Verboo Code with Pi

With Node.js and npm installed, install Pi. Your Verboo Code account needs access to the model you want to use.

npm install -g --ignore-scripts @earendil-works/pi-coding-agent

Create a key in the dashboard and replace <VERBOO_API_KEY> below. Export the variable in the same terminal where you will launch the CLI. The configuration reads this variable for authentication, keeping your key out of the JSON file.

export VERBOO_API_KEY="<VERBOO_API_KEY>"

On Windows, set the variable with PowerShell:

$env:VERBOO_API_KEY = "<VERBOO_API_KEY>"

Create ~/.pi/agent/models.json and its parent folders if they do not exist. If the file already exists, add the verboo entry to the providers object while preserving your current configuration.

{
  "providers": {
    "verboo": {
      "baseUrl": "https://code.verboo.ai/router/v1",
      "api": "openai-completions",
      "apiKey": "$VERBOO_API_KEY",
      "authHeader": true,
      "models": [
        {
          "id": "<MODEL_ID>"
        }
      ]
    }
  }
}

In Pi, openai-completions selects Chat Completions. The apiKey variable uses a $ prefix. Adjust contextWindow and maxTokens to your model if needed; Pi defaults may exceed its limits. On Windows, use %USERPROFILE%/.pi/agent/models.json.

Open a terminal in your project folder and start the CLI with your configured model:

pi --provider verboo --model "<MODEL_ID>"

Inside Pi, use /model to select models from the verboo provider. Reopen the picker to reload file changes.

To check available IDs and key access, query the API in Bash or Zsh:

curl https://code.verboo.ai/router/v1/models \
  -H "Authorization: Bearer $VERBOO_API_KEY"

For a 401 response, check the key and environment variable. For 403, check model access in your plan. If the model is missing, check its ID and provider configuration.