Last updated: 2026-06-28 00:00 Asia/Singapore
Use:
https://apihub.agnes-ai.com/v1
Agnes AI is designed for OpenAI-compatible integrations. Chat workflows use:
POST /v1/chat/completions
| Workflow | Recommended Model |
|---|---|
| Coding, reasoning, tool calling, and agent workflows | agnes-2.0-flash |
| Text-to-image and image editing | agnes-image-2.1-flash |
| Fast image generation | agnes-image-2.0-flash |
| Text-to-video and image-to-video | agnes-video-v2.0 |
See MODEL_CATALOG.md for the current reference RPM and subscription quota tables. See TOKEN_PLAN_FAQ.md for access types, Token Plan quotas, and API key limit-pool behavior.
Limits may change over time. Always confirm production-critical values in the official Agnes AI platform console.
The public reference video RPM values were updated:
| User type | Public request RPM | Actual executable RPM |
|---|---|---|
| Free / default | 2 | 1 |
| Enterprise | 2 | 2 |
| Token Plan | 6 | 5 |
Token Plan video subscription quota remains 500 seconds per day.
The current public reference value is 512K context and 64K max output.
Use the returned video_id:
GET https://apihub.agnes-ai.com/agnesapi?video_id=<VIDEO_ID>
Do not use task_id for current video result polling unless a legacy integration specifically documents that workflow.
Include the model, endpoint, SDK or client, sanitized request body, error code, expected behavior, actual behavior, and whether the issue is reproducible.
Do not include API keys, bearer tokens, private logs, or customer data.