Developer Documentation
Use one API key to call text, image, and video models through their native protocols.
HopsAPI is a native multi-protocol AI API gateway. Discover capabilities from the model list, then call the protocol and endpoint advertised for each model. The gateway handles authentication, model-access checks, usage metering, and response forwarding.
HopsAPI does not force every model into one protocol. Claude, Gemini, OpenAI-compatible text models, and media models should use their respective interfaces.
1. Choose a capability
| Capability | Models | Interface | Start here |
|---|---|---|---|
| Chat generation | GPT, Kimi, MiniMax, DeepSeek, Grok | OpenAI Chat Completions | OpenAI Chat |
| Responses | text models that return responses_endpoint | OpenAI Responses | Responses API |
| Claude | Claude family | Anthropic Messages | Anthropic |
| Gemini | Gemini family | Google generateContent | Gemini |
| Video generation | Grok Imagine Video | Async media tasks | Grok Imagine Video |
| Image generation | Gemini image, Nano Banana Pro, Nano Banana 2, GPT Image 2, GPT Image 2.5 | Async media tasks; GPT Image 2 and GPT Image 2.5 also have OpenAI-compatible image endpoints | Gemini image, Nano Banana Pro, Nano Banana 2, GPT Image 2, or GPT Image 2.5 |
2. Standard integration flow
- Create and securely store an API key in the console.
- Call
GET /v1/modelsto list the models and endpoints available to the key. - Select a protocol from the model's
format,endpoint, and optionalresponses_endpoint. - Follow the corresponding protocol guide and handle its native response.
GET
/v1/modelsList the models, protocol formats, and endpoints available to the current API key.
3. Gateway capabilities
| One key, many models | Authenticate Claude, GPT, and image/video models without separate upstream accounts. |
| Native protocols | Use Chat Completions, Responses, Anthropic Messages, Gemini generateContent, and async media tasks. |
| Streaming or one-shot | Relevant text endpoints support native SSE streaming and non-streaming responses. |
| Independent usage controls | Usage is metered per API key, with optional spend limits. |
| Media-task protection | Fixed-price media tasks reserve the configured price at creation; only successful completion settles a charge, while failure or cancellation releases the reservation. |
4. Recommended reading order
- Quick start — make your first text request.
- Protocols & endpoints — understand why model families use different interfaces.
- List models — inspect the capability-discovery response.
- Video generation: Grok Imagine Video.
- Image generation: Gemini image, Nano Banana Pro, Nano Banana 2, GPT Image 2, and GPT Image 2.5.
- Error handling — handle errors from text, media, and compatible image endpoints.