Glossary · AI
OpenAI-compatible endpoint
Also called: OpenAI-compatible API, custom base URL, chat completions compatible
An OpenAI-compatible endpoint is an AI API that accepts requests and returns responses in the same format as OpenAI's API, so software written for OpenAI can call other providers, gateways or self-hosted models by changing the base URL and key.
OpenAI's chat completions request format became a de facto standard. Many providers, routers and open-source model servers accept the same shape: a base URL, a bearer key, a model name and a list of messages. If a tool lets you set the base URL, it can talk to any of them.
What people use it for
- Model routers and gateways that put many providers behind one key and handle fallbacks.
- Inference providers hosting open-weight models.
- Self-hosted models running on your own GPUs or a private cloud, exposed through a compatible server.
- Company AI gateways that add logging, redaction or budget controls in front of providers.
Compatibility isn't identical
"Compatible" usually covers basic chat requests. Features like JSON mode, tool calling, streaming and token usage reporting vary between servers, and some only implement a subset. Test the exact features your workflow relies on, including how errors and rate limits are reported, before you depend on it.
Security note
An endpoint that's reachable from the public internet needs authentication and TLS; a model server bound to a private network shouldn't be exposed just so a SaaS tool can reach it. Put a gateway with its own key in front of it instead.
In Fluxo Flow
"Any OpenAI-compatible endpoint" is one of the 14 provider options: enter the base URL and key, and pick the model per step. It must be reachable at a public, secured https address; local and private network addresses are blocked. That's also how a self-hosted or niche model plugs in.
Read next
Related terms
- BYOK (bring your own key)BYOK, or bring your own key, means a software product runs its AI features on an API key from your own AI provider account, so usage is billed to you directly by the provider instead of being resold through the product.
- LLMAn LLM, or large language model, is a neural network trained on large amounts of text that generates text by repeatedly predicting the next token, which lets it summarize, classify, extract and draft.
- Rate limitA rate limit is a cap an API puts on how many requests a client may make in a time window; requests over the limit are refused, usually with HTTP status 429 Too Many Requests.
- Polling vs webhooksPolling means your system asks another system for new data on a fixed schedule; webhooks mean the other system pushes each event to you the moment it happens.
- OAuth 2.0OAuth 2.0 is an authorization standard that lets you grant one application limited, revocable access to your account in another service, without giving it your password, by issuing it access tokens with specific scopes.