Your AI, your choice
Different companies have different rules about where data may be processed and which AI vendor they trust. Rather than lock you into one, the Layer Assistant lets an administrator choose the AI service it uses, in Settings > AI Chatbot.
The options at a glance
OpenAI - the quickest to set up; you supply an OpenAI key.
Azure AI Foundry - OpenAI models running inside your own Microsoft and Azure tenancy, for the tightest data control.
Claude - Anthropic's models.
Self-hosted - an OpenAI-compatible model on your own hardware, so nothing leaves your network.
OpenAI or Claude
Choose the provider, paste your key in the AI API key box, and you are ready. These are the simplest to get going with.
Azure AI Foundry, for the tightest control
If your organisation already runs Microsoft 365 and Azure, Azure AI Foundry lets you use OpenAI models inside your own tenancy. Prompts and answers stay within the Azure environment you govern, under your own region, data-retention and compliance settings, which is often exactly what a security or data-protection team wants to see.
To set it up:
Set Provider to Azure AI Foundry.
Paste your Azure key in AI API key.
In Completion URL, paste the full Target URI from the Foundry portal (it already contains the deployment and version), or just your resource endpoint such as https://your-resource.openai.azure.com and put the deployment name in Model.
Azure AI Foundry is billed through your own Azure account, so its usage counts towards the assistant's monthly budget just like the other cloud providers.
Self-hosted
If you run your own OpenAI-compatible model, put its address in Completion URL and leave the provider as OpenAI. The assistant will talk to it instead of the cloud, and self-hosted turns are not billed.
Pick a model (optional)
Leave Model blank to use a sensible default, or name a specific model such as gpt-4o. For Azure AI Foundry, this box holds your deployment name unless it is already part of the Completion URL.
Set a spend cap
You can set a monthly budget so the cost of the assistant stays bounded. See Turn on the Layer Assistant for that and the other settings.