---
title: API overview
description: Choose a compatible API format, configure authentication, and discover the exact model IDs available to your key.
lastUpdated: 2026-09-08
---

Infistar offers a shared account and API key across multiple AI models. Keep the request format your application already uses and select a model that supports that endpoint.

## Base URL

For OpenAI-compatible clients, use:

```text
https://infistar.ai/v1
```

For networks in mainland China, use `https://infistar.cc/v1`. Both addresses use the same account and API key. Configure one base URL per client; do not append `/v1` twice. Claude Code uses the site origin without `/v1`, as explained in its [setup guide](/en/client-integrations/claude-code).

## Authentication

Create a key in **Console → Token Management** and send it in the request header:

```text
Authorization: Bearer YOUR_API_KEY
```

Keep the key on your server. See [Authentication and permissions](/en/integration-guides/authentication-permissions).

## Discover available models

Call `GET /v1/models` with the key you will use. Copy the exact returned model ID; a display name, plan name, or supplier name is not a model ID. A listed model does not necessarily support every endpoint. Check its current interface types in [Model Square](https://infistar.ai/pricing).

## Select an endpoint

| Workflow | Endpoint | English guide |
| --- | --- | --- |
| Available models | `GET /v1/models` | [Models](/en/api-reference/models) |
| Chat messages | `POST /v1/chat/completions` | [Chat Completions](/en/api-reference/chat-completions) |
| Responses and structured input | `POST /v1/responses` | [Responses](/en/api-reference/responses-api) |
| Anthropic Messages | `POST /v1/messages` | [Messages](/en/api-reference/claude-format-api) |
| Image generation or editing | `POST /v1/images/generations`, `POST /v1/images/edits` | [Images](/en/api-reference/image-generation) |
| Video jobs and results | `POST /v1/videos`, `GET /v1/videos/{task_id}` | [Videos](/en/api-reference/video-generation) |

Gemini, embeddings, reranking, realtime audio, speech, and transcription depend on the selected model and endpoint. The broader [compatibility matrix (Chinese)](/api-compatibility-matrix) provide the remaining interfaces.

Start with a small non-streaming request. Then test only the streaming, tools, media, or other capabilities your application actually requires. Check the response and usage log before increasing traffic.