Native Realtime Model Integration

ChatBotKit now integrates realtime models natively. Rather than treating realtime as a separate, limited mode, a realtime model behaves like any other model on the platform - assigned to a bot, deployed the same way, and fully interoperable with the rest of your stack.

The same model works in both chat and voice mode. You can hold a text conversation or a live, low-latency voice conversation with the same bot, the same configuration, and the same behaviour behind it. There is no separate bot type and no parallel setup to maintain.

Most importantly, realtime models have full access to everything your agents already use. The complete skillset engine is available, so the model can call tools, search your datasets, read and write memory, use credentials, and reach into your internal systems and connected integrations mid-conversation - in text or in voice. Realtime is no longer a constrained experience; it is a first-class way to run a fully capable agent.

This unlocks genuinely interactive voice and chat agents that can act, not just talk - answering questions grounded in your data, completing workflows across connected services, and remembering context from one conversation to the next.

Native realtime support is available today. Assign a realtime model to a bot and use it in chat or voice with the full power of the ChatBotKit agent stack behind it.