Introducing an OpenAI-Compatible API

We're adding an OpenAI-compatible chat completions endpoint to the ChatBotKit API. If you already work with the OpenAI SDKs, or with any tool that speaks the OpenAI Chat Completions format, you can now point it at ChatBotKit with only a couple of lines of configuration.

How It Works

Set your client's base URL to the ChatBotKit OpenAI-compatible endpoint and authenticate with a ChatBotKit secret key, just as you would with an OpenAI API key.

Selecting a Model or Agent

The model field accepts a simple selector so you can target exactly what you want:

• model/name=<model> - talk to a specific model directly
• bot/id=<id> - use one of your configured agents, along with its backstory, datasets and skillsets

Any system prompt you pass is applied on top of the selected target, so you can layer instructions without overriding an agent's existing configuration.

Systems as Models

Because bot/id=... can point at a full agent, a single "model" can quietly bundle a backstory, datasets, skillsets, tools, built-in search, and even a team of coordinated agents - similar to how Perplexity packages search into what looks like an ordinary model. To the calling client it's just a model; behind the scenes it's an entire ChatBotKit system. Learn more on the OpenAI-Compatible API feature page.

What's Supported

• Both standard and streaming responses (server-sent events)
• Tool and function calling, in the current tools format and the legacy functions format
• Standard sampling parameters - parameters we don't map are accepted and ignored, so existing client code keeps working

Why This Matters

The goal is compatibility, not lock-in. Existing OpenAI-based applications, libraries and frameworks can adopt ChatBotKit with a configuration change rather than a rewrite, while still benefiting from ChatBotKit agents, knowledge and tooling under the hood.

The endpoint is documented in our API reference alongside the rest of the ChatBotKit API.