Configuration

Model configuration

The TypeScript build skips project-root .env by default. AgentHub or scripts inject provider keys through SDK env or controlled shell environments. The Gateway preset writes provider, model, and base URL, while the caller owns the API key.

$env:OPENAI_API_KEY = "your-openai-key"
$env:TOKENDANCE_GATEWAY_API_KEY = "your-gateway-key"
tokendance gateway init --model deepseek-v4-pro
tokendance config validate --json

Config facade

The SDK/CLI config facade reads and writes the allowlisted provider, model, and permissionMode fields, then filters secret-like fields such as apiKey and token. config validate --json returns readiness, missing inputs, required key envs, selected key env, and base URL source for AgentHub settings or onboarding.

tokendance config --json
tokendance config validate --json
tokendance config set --json provider openai-chat-completions model deepseek-v4-pro permission-mode safe