Docs menu

using the panel

Models and effort

extynt is bring-your-own-model. You choose the provider and pay them directly.

Set up a model

  1. Open Settings › Model in the panel.
  2. Choose a service, then a model from its list, or pick Custom model ID….
  3. Paste your API key. It stays on your computer.
  4. Press Save. extynt then tests the model and shows whether it answered. Press Test any time to check it again.

Supported services include Anthropic, OpenAI, DeepSeek, xAI, OpenRouter, any OpenAI-compatible or Anthropic-compatible endpoint, and Ollama for local models with no key.

Fallback model

Optionally add a fallback under Fallback model. It is only used when the first model fails, and the chat tells you when it takes over.

Reasoning effort

The effort button next to Send switches provider, model and reasoning effort on the fly. Only the levels your model supports are offered. Higher effort is slower and costs more. Use it for large or fiddly changes.

Spending guard

Settings › Behaviour › Pause after heavy use is on by default. It stops a runaway loop from burning through your API budget.

Using Claude or Codex instead

You can skip the panel’s model and drive extynt from your AI client over MCP, on the Claude or Codex account you already have. See How MCP works.