You'll see “OpenAI-compatible” everywhere: in editors, chat apps, automation tools and libraries. It means something specific and useful.
The short version
The OpenAI Chat Completions API became the de-facto standard way to talk to language models. An OpenAI-compatible API accepts the same request shape at the same paths (like /v1/chat/completions) and returns the same response shape. Any software written for OpenAI can therefore talk to it by changing two settings:
- Base URL:
https://devsaio.lol/v1 - API key: your DEVSAIO key
What's usually covered
| Feature | Path |
|---|---|
| Chat (text and images in, text out) | /v1/chat/completions |
| Streaming | same path, "stream": true |
| Tool / function calling | tools field |
| Model list | /v1/models |
| Embeddings | /v1/embeddings |
| Image generation | /v1/images/generations |
The model name decides which lab answers. Ask for claude-sonnet-4-6 and Anthropic's model responds; ask for gemini-3.5-flash and Google's does, but your code doesn't change.
A quick test
curl https://devsaio.lol/v1/models -H "Authorization: Bearer $DEVSAIO_API_KEY"If that returns a list of models, any OpenAI-compatible tool will work with the same two settings.
Where to put the settings
- Python / Node SDKs:
base_url/baseURLwhen creating the client. - Environment variables: many tools read
OPENAI_BASE_URLandOPENAI_API_KEY. - Apps: look for “Custom OpenAI endpoint”, “OpenAI-compatible provider” or “API base”.
Note: Some apps expect the base URL with
/v1at the end and some without. If you get a 404, try the other form.
What about Anthropic's own format?
Tools built for Claude (like Claude Code or the Anthropic SDK) use the Anthropic Messages format instead. DEVSAIO supports that too: see using the Anthropic-compatible endpoint.
Claude, GPT, Gemini, DeepSeek, image and video models through one OpenAI- and Anthropic-compatible endpoint. Prepaid, pay only for what you use.