Unified AI API — One Key for 100+ Models
Unified AI API
A unified AI API is a single authenticated surface that routes requests to many underlying model providers without forcing integrators to learn separate SDKs, auth schemes, or polling conventions. ModKie implements this pattern with one API key, one async job contract (createTask + recordInfo), and a catalog of 100+ models spanning Google Veo, Kling, Runway, Suno, OpenAI-family image models, and self-registered RunPod deployments.
Developers discover models in the market, validate payloads in the Playground, then promote the same JSON to production. The gateway normalizes provider differences behind Kie-compatible task IDs and exposes per-model creditCost in GET /models. OpenAI-style chat remains available at /api/gateway for conversational workloads that do not use the async job pipeline.
Provider coverage (sample)
| Provider hub | Example models | Task types |
|---|---|---|
| Veo 3.1, Nano Banana 2 | Video, image | |
| Kling | Kling 3.0 | Video |
| Runway | Gen-4 Turbo | Video |
| Suno | Suno V4 | Music |
| OpenAI | GPT Image 2 | Image |
| Custom | Your RunPod model | As registered |
When to use createTask vs chat gateway
- createTask: Video, image, music, and long-running generation with explicit
taskIdpolling. - Chat gateway: Streaming or single-shot LLM completions compatible with OpenAI client libraries.
FAQ
How do I get started?
Follow Getting Started and Market Quickstart.
Are failed jobs charged?
No. Credits deduct only when recordInfo reports success.
Last updated: 2026-06-26