Configuration
Settings file (CLI / agent)
TerminaI uses the same settings file layout as the upstream Gemini CLI.
- Default path:
~/.terminai/settings.json
Common options:
security.approvalPin(string, 6 digits)- Used for Level C approvals (default:
"000000"). Example:"123456"
- Used for Level C approvals (default:
security.approvalMode(string: "safe" | "prompt" | "yolo")- Controls approval ladder behavior (default:
"prompt").
- Controls approval ladder behavior (default:
previewMode(boolean)- Enables experimental features (default:
false).
- Enables experimental features (default:
provider(string: "gemini" | "ollama")- AI model provider (default:
"gemini").
- AI model provider (default:
voice.enabled(boolean)- Enables CLI spoken replies (TTS).
voice.pushToTalk.key(string)- CLI key for voice controls (commonly
space).
- CLI key for voice controls (commonly
voice.spokenReply.maxWords(number)- Caps how much text is spoken per assistant turn.
Environment variables:
GEMINI_API_KEY- Uses API-key auth instead of the OAuth browser flow.
Web Remote (A2A) token
When you start the agent with --web-remote, the CLI prints (or stores) a token
used by clients.
- Rotate token (prints a new token):
terminai --web-remote-rotate-token - Start server with a pinned port:
terminai --web-remote --web-remote-port 41242
Desktop app settings
Desktop stores its own UI settings locally (agent URL/token, workspace path,
voice toggle/volume). These settings do not replace the agent’s
~/.terminai/settings.json.