Understand models and channels
Understand providers, channels, models, and the endpoint your application uses.
Four connection concepts
- Provider
- An upstream source of model services.
- Channel
- A connection configuration for an upstream service, including its address and required credentials.
- Model
- The model resource an application requests; it must match the channel and supported protocol.
- Gateway endpoint
- The address applications use to reach BOS API, managed separately from upstream addresses.
API-key and official subscription channels
Channels can use an API key / upstream proxy or connect ChatGPT / Codex, Claude, Gemini CLI / Code Assist, and Antigravity through OAuth. Official subscriptions pin the endpoint and protocol, and still require configured models and organization / key authorization. A model list or preset name does not establish account access.
Follow the configuration relationships
An application uses a gateway key to request a model. The gateway organizes forwarding based on model, authorization, and channel configuration. Use model identifiers confirmed by the administrator; do not assume how provider model names are mapped.
Check protocol compatibility
An available model does not imply that every calling interface is supported. Confirm the protocol, client version, request parameters, streaming behavior, and error handling before integrating.
This site does not publish unverified provider counts or protocol coverage. Consult the verified compatibility information for your deployment version.
Match capabilities and procedures to your deployment version. Read the documentation scope