Standard API setup: OpenAI / Anthropic
If your software accepts a custom API URL, choose one of the supported protocols below and enter API URL, API key and model ID. No client code changes are needed. The same AI526 key works with both protocols.
OpenAI standard setup
| Client field | Value |
|---|---|
| Protocol / Provider | OpenAI / OpenAI Compatible, Chat Completions |
| API URL / Base URL | https://api.ai526.com/v1 |
| API Key | Your AI526 key from key management |
| Model | A public chat model ID from the catalog, such as deepseek-v4.1-flash-1017 |
Most clients append /chat/completions. If the field requires a full endpoint URL, use https://api.ai526.com/v1/chat/completions. If the client appends /v1/chat/completions itself, use https://api.ai526.com to avoid a duplicate /v1.
HTTP authentication is Authorization: Bearer YOUR_AI526_API_KEY. Chat and streaming are supported; check the selected model for tools and image input capabilities.
Anthropic standard setup
| Client field | Value |
|---|---|
| Protocol / Provider | Anthropic / Claude, Messages |
| API URL / Base URL | https://api.ai526.com (without /v1) |
| API Key | The same AI526 key; no separate official key is needed |
| Model | A public chat model ID with the capabilities you need, such as deepseek-v4.1-flash-1017 |
The Anthropic SDK appends /v1/messages. If a full endpoint URL is required, use https://api.ai526.com/v1/messages. If your client appends only /messages, use https://api.ai526.com/v1 as its base URL.
HTTP requests use x-api-key: YOUR_AI526_API_KEY and the compatible version header anthropic-version: 2023-06-01. The body includes model, max_tokens and messages; normal clients construct these fields automatically.
Match the protocol and model
A protocol is a communication format, separate from a model's original company. Use chat models with these interfaces. Image and video models require their own endpoints and cannot serve as a chat assistant. See the OpenAI endpoints and Anthropic endpoints.
OpenAI Responses (/v1/responses) is currently unavailable. Responses-only clients cannot use this setup yet; see Codex compatibility. A client that hardcodes an official host also needs a custom URL setting or administrator configuration.
Check your connection
Save the settings, send a short message and review request history. For 401, check your key; for 404, check the path and model ID. Check key model restrictions and account status if a model is unavailable. Provider routing is automatic by default.
Protocol references
AI526 compatibility is limited to its published endpoints and model capabilities.