This article explains what runs behind an AI agent, how you pay for it, and what to do if you need a specific provider or your own key.
HueChat's AI is built on a provider-agnostic layer. Depending on the task (answering, embedding your knowledge, retrieval, evaluation and safety checks), HueChat may use models from OpenAI, Anthropic and Google, and supporting services such as Groq, DeepSeek and Jina AI. HueChat routes each step to a suitable model tier, so a simple FAQ answer and a complex booking conversation do not cost the same.
You do not pick a model per agent. You configure behaviour (Instructions, knowledge, handoff) and HueChat handles the models.
Every plan, including Free, includes a monthly allowance of AI Actions. Each answer the AI sends uses some AI Actions, based on how much work it does. The AI Actions page shows:
Test messages in the agent editor are free within the test limits and do not use AI Actions. See the pricing page for each plan's allowance.
Screenshot: the AI Actions page with the current balance and the Buy More AI Actions button
Provider keys are managed by HueChat. There is currently no self-service field in the dashboard to enter your own OpenAI, Anthropic or Google key for the AI agents, and usage is metered through AI Actions rather than through your provider account.
If your organisation requires a dedicated key, a specific provider, or a data-processing arrangement with one provider, contact support@huechat.ai. Enterprise arrangements can be discussed case by case. Exact options may differ by plan.
Note the API Key field you will see inside an agent belongs to the Booking Integration and Webhooks sections. It authenticates HueChat to your own systems (for example your booking API); it is not a model provider key.
If you want to run your own models or automation on top of HueChat, use the public API and webhooks documented in the Developer Hub (developers.huechat.ai). API access and webhooks are available from the Growth plan.