Settings

Provider keys, model pricing, and the defaults used when you start a new discussion. Everything is stored locally on your device.

Setup incomplete

Configured: 0

Enabled: 0

Validated: 0

Provider health

You need at least two enabled models to run a multi-model discussion cleanly.

API keys saved

Providers with stored credentials

Enabled models

Available on the home screen

Tested models

Models with a successful connection test

Add and enable at least two providers, then choose a default judge model in the Defaults tab.

API keys

Connect any OpenAI-API-compatible endpoint — a model you run locally (Ollama, LM Studio) or a hosted server. For local Ollama, base URL http://localhost:11434/v1 and model gemma4:12b. Ollama must also allow requests from the browser: set OLLAMA_ORIGINS=* before starting it (PowerShell: $env:OLLAMA_ORIGINS="*"; ollama serve), otherwise the browser blocks the connection with a CORS error. Use the Supported inputs toggles to declare which media types the model accepts. If you just want to use a newer OpenRouter model, add its id on the OpenRouter tab instead of creating a custom endpoint here.

Add a custom model

Display name

Model id

Base URL

API key (optional)

Supported inputs

Text is always supported. Enable what this endpoint accepts — image and document attachments are sent over the OpenAI-compatible API.

Declared tool support

Tool support is not auto-detected. Enable a tool only when this endpoint actually implements it; these switches are your explicit declaration of endpoint support.

Compatible API transports

Chat Completions is the conservative default. Declare Responses API only if this endpoint implements the OpenAI-compatible Responses path.