Managing providers

The Chat Agent works with your own AI provider API keys. You connect each provider once inside the app and all of its models become instantly available in every chat. You only need to connect the providers you intend to use — unconnected providers are hidden from the model selector. For information on choosing a model after connecting a provider, see Selecting models.

Adding a provider

To connect a new provider:

1. Open Settings → Providers from the sidebar or the user menu.
2. Click Add provider.
3. Choose the provider from the list (OpenAI, Anthropic, Azure OpenAI, Google Gemini, Ollama, or OpenAI-compatible).
4. Enter the required credentials for that provider (see each section below).
5. Click Save. The app validates the credentials immediately and the provider's models appear in the model selector right away.

Supported providers

OpenAI

Paste your OpenAI API key (beginning with sk-) into the API key field. Once saved, models such as GPT-4o, GPT-4o mini, and GPT-4 Turbo will appear in the model selector. All OpenAI models support tool use and streaming.

Anthropic

Paste your Anthropic API key (beginning with sk-ant-) into the API key field. Claude 3 Opus, Claude 3.5 Sonnet, Claude 3 Haiku, and subsequent releases will appear in the model selector. All Claude models support tool use and streaming.

Azure OpenAI

Azure OpenAI requires three fields in the Add provider dialog:

• Endpoint URL — your Azure OpenAI resource endpoint, for example https://my-resource.openai.azure.com.
• Deployment name — the name of the deployment you want to use as the default (e.g. gpt-4o-prod). You can select other deployments from the model selector per chat.
• API key — the API key for your Azure OpenAI resource.

Any model available in your Azure subscription — GPT-4o, GPT-4 Turbo, GPT-4o mini — will appear once the connection is validated.

Google Gemini

Paste your Google AI Studio API key (beginning with AIzaSy) into the API key field. Gemini 1.5 Pro (large context, complex reasoning) and Gemini 1.5 Flash (faster, lower cost) will appear in the model selector. Both support tool use and streaming.

Ollama and OpenAI-compatible endpoints

Any server that implements the OpenAI chat completions API — including Ollama and similar services — can be connected as an OpenAI-compatible provider. In the Add provider dialog:

• Endpoint URL — the base URL of your running server, such as http://my-server:11434 for Ollama or the URL of any other compatible endpoint.
• API key — leave blank if the server does not require authentication, or enter the key if it does.
• Model name — the exact model identifier the server expects (check the server's own documentation for the correct string).

The model selector will list whichever models the endpoint reports as available. Tool use support depends on the server and the model it is running.

Editing and removing a provider

To update the credentials for an existing provider, open Settings → Providers, find the provider in the list, and click Edit. You can update the API key, endpoint URL, or any other field, then click Save to re-validate and apply the changes.

To disconnect a provider, click Remove next to it. The provider and all of its models will be removed from the model selector. Any chats that used a model from that provider will retain their history, but you will not be able to continue them with the same model until the provider is reconnected.