Skip to main content
By default, Wraith agents use a hosted Gemini model. You can replace it with OpenAI, Claude, or your own Gemini key to reduce cost or use a specific model.

Configuration

Pass the ai option when creating the Wraith client:
All agents created from this client use your specified AI provider.

Supported Providers

OpenAI

Claude

Your Own Gemini Key

How It Works

The AI configuration is passed to the TEE server via HTTP headers on every request: The TEE server:
  1. Receives your chat message
  2. Builds the system prompt with agent identity and tools
  3. Sends the conversation to your specified AI provider
  4. Executes any tool calls via the chain connector
  5. Returns the final response
Your AI key is only used server-side in the TEE. It’s never stored — used for the duration of the request only.

Default vs. BYOM

Tool Compatibility

All tools work identically regardless of the AI provider. The tool declarations are adapted to each provider’s function-calling format by the TEE server. Your code doesn’t change.